Skip to main content

Full text of "Some acoustic and perceptual correlates of the modal and falsetto registers"

See other formats


SOME ACOUSTIC AND PERCEPTUAL CORRELATES 
OF THE MODAL AND FALSETTO REGISTERS 



By 
RAYMOND H. COLTON 



A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF 

THE UNIVERSITY OF FLORIDA 

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE 

DEGREE OF DOCTOR OF PHILOSOPHY 



UNIVERSITY OF FLORIDA 
1969 



U OF F LIBRARIES 




.....-■ .■ :- 



■,-■ ■■■■ y ^ | tf 



To my wife Gloria and daughter Laurie 



W0H:A\ i; ''L\ 



ACKNOWLEDGMENTS 

The author gratefully acknowledges the helpful 
guidance, constructive criticism, and patience of Dr. 
Harry Hollien throughout the course of this study as well 
as the writer's graduate career at the University. The 
author is especially appreciative of Dr. Hollien ' s efforts 
to provide a balanced and, thorough program of research train- 
ing with which the author can confidently initiate and conduct 
his own research in the future. 

Grateful acknowledgment is made to the writer's 
supervisory committee consisting of Drs. Dew, Paige, Moore, 
Teas, Algeo, and Keister; all of whom have made many con- 
structive comments about this study. A special note of 
appreciation is also made to all of the members of the Com- 
munication Sciences Laboratory; both faculty and graduate 
students who .listened to each research proposal carefully, 
and offered many constructive suggestions and alternate 
approaches to the specific research problem at hand. 

Special appreciation is reserved for the author's 
wife Gloria, who always supported his endeavors and helped 
him during several portions of this study and during the 
writing of t h e m anuscr i p t . 



m 



TABLE OF CONTENTS 

ACKNOWLEDGMENTS ' iii 

LIST OF TABLES v 

LIST OF FIGURES . viii 

CHAPTER 

I INTRODUCTION - 1 

II PROCEDURE 19 

III RESULTS AND DISCUSSION 30 

IV SUMMARY AND CONCLUSIONS 8 4 

APPENDIX A 88 

APPENDIX B ; 9 2 

APPENDIX C 97 

APPENDIX D 103 

BIBLIOGRAPHY 107 

BIOGRAPHICAL DATA Ill 



IV 



LIST OF TABLES 



Tabl e 
1 



Mean phonational range of singers and 

naive subject in the modal and falsetto 
registers 



Differences between naive and singer 
subjects at each register boundary, 
the semitone overlap between the 
modal and falsetto registers, and the 
total phonational range 



Mean phonational ranges of the modal and 
falsetto registers for naive and singer 
subjects on each of three phonational 
range screening sessions .... 



Vocal intensity levels in the modal and 
falsetto registers for naive and singer 
subjects phonating at three frequency 
levels within the region of register 
overlap 



Pag! 



32 



33 



39 



44 



Summary of the analysis of variance per- 
formed on vocal intensities measured 
at each of three frequency levels in 
the modal and falsetto registers for 
the naive and singer subjects . . . . 



Proportion of correct responses as same- 
different for ph on at ions produced in the 
modal and falsetto registers at th: 
frequency levels 



r ee 



Summary of the analysis of variance per- 
formed on the arcsine transformations of 
the proportion of correct responses in 
the p a i r e d c o m p a r i s o n p e r e e p t u a 1 pro- 
cedure 



4S 



53 



54 



Distribution of modal and falsetto 
responses on the first and second 
presentation of samples produced by 
naive and singer subjects in the 
forced choice perceptual procedure 



60 



■ ; ;fc:- 



Distribution of modal and falsetto 
responses on the first and second 
presentation of samples produced 
by naive and singer subjects in the 
forced choice perceptual procedure 



61 



10 



11 



1 2 



13 



14 



Summary of the analysis of variance 
performed on the arcsine trans- 
formations of the proportion of 
correct responses for three frequency 
levels and two subject groups in the 
forced choice perceptual procedure . . 

Proportion of correct categorizations of 
the modal and falsetto register pho- 
nations produced by singers and naive 
subjects at each of three frequency 
levels within the range of register 
overlap in the forced choice cate- 
gorization procedure 



Distribution of modal and falsetto respon- 
' ses on the first and second presenta- 
tions of samples produced by naive 
and singer subjects in the free choice 
perceptual procedure 



Summary of the analysis of variance per- 
formed on the first responses to modal 
and falsetto register pronations pro- 
duced at three frequencies by naive 
and singer subjects in the free choice 
perceptual procedure 



Summary of the analysis of variance per- 
formed on the second responses to modal 
and falsetto phonations produced at 
three frequencies by naive and singer 
subjects in the free choice perceptual 
procedure 



63 



66 



69 



71 



72 



VI 



15 Proportion of correct responses for p h o - 

nations produced in the modal and falsetto 
registers by naive and singer subjects at 
three frequency levels in the free choice 
perceptual procedure 



76 



16 Summary of the analysis of variance per- 
formed on the number of partials mea- 
sured in the phonations produced by 
naive and singer subjects at each 
frequency level 



82 



17 Summary of the analysis of variance for 
simple effects conducted on the 
register/frequency /intensity inter- 
action for the vocal intensity data 



98 



18 Summary of analysis of variance for 

simple effects conducted on register/ 
intensity interaction for the vocal 
intensity data 



99 



19 Summary of the analysis of variance 

for simple effects conducted on the 
observer group/ subject interaction 
(GS) and the subject group/frequency 
(SF) interaction for the paired com- 
parison procedure 



100 



20 Summary of the analysis of variance 

for simple effects conducted on the 
observer groups/subject groups/ 
frequency (GSF) interaction for the 
forced choice perceptual procedure , 



101 



21 Summary of the analysis of varianc for 
simple effects conducted on the sub- 
ject group/regi s ter/ frequency 
interaction (SRF) and the register/ 
frequency interaction (RF) of the 
first response data in the free 
choice procedure 



102 



VI 1 



15 Proportion of correct responses for pho- 
nal ions produced in the modal and falsetto 
r egis t e r s b y n a i v e a n d singer subjects a t 
t h r e e f r e q u e n c y 1 e v e 1 s in the free choice 
perceptual procedure 



76 



Summary of the analysis of variance per- 
formed on the number of harmonics mea- 
sured in the phonations produced by 
naive and singer subjects at each 
frequency 1 evel 



82 



3 7 Summary of the analysis of variance for 
simple effects conducted on the 
register/frequency/intensity inter- 
action for the vocal intensity data 



98 



IS Summary of analysis of variance for 

simple effects conducted on register 
intensity interaction for the vocal 
intensity data 



99 



Summary of the analysis of variance 
for simple effects conducted on the 
observer group/ subject interaction 
(GS) and the subject group/frequency 
(SF) interaction for the paired com- 
parison procedure 



100 



20 



21 



Summary of the analysis of variance 
for simple effects conducted on the 
observer groups/subject groups/ 
frequency (GSF) interaction for the 
forced choice perceptual procedure . 

Summary of the analysis of varianc for 

:ond u c t e d o n t h e s u b - 



simple effects 

ject 

i n t e r actio 



g r o u p/ r e gister / f re q u e n c 



y 



( S R F ) 



of the 



i a t h e r egi ster/ 
frequency interaction (RF) 
first res p o n s e d a t a i n t h c f re 
choice proc e d uro 



103 



102 



vi i 



LIST OF FIGURES 



Figure 



Mean phonational ranges of naive 
and singer subjects in the 
modal and falsetto registers. 



Page 



37 



Mean vocal intensity of naive 

subjects at the frequency levels 
of the range of register over- 
lap in the modal and falsetto 
registers 



45 



Mean vocal intensity of singers 
at each frequency level in the 
modal and falsetto registers 



46 



Proportion of correct responses 
in the paired comparison per- 
ceptual procedure to phonations 
produced at three frequency levels 
by the naive and singer subjects . . 57 

Proportion of correct responses by the 
naive and experienced observers 
who rated phonations produced by 
the naive and singer subjects at 
each of three frequency levels 
within the range of register overlap 
in the forced choice perceptual 
procedure 65 



Proportion of correct responses to 
modal and falsetto register phona- 
tions produced by naive and singer 
subjects at the three frequency 
levels of the range of register 
overlap. Data are from the -first 
responses to the stimuli . 



73 



Proportion of correct responses to 
modal and falsetto register pho- 
nations produced by naive and 
singer subjects at three frequency 
levels within the range of register 
overlap in the free choice perceptual 



procedure. Data are from the second 
presentation of the stimuli .... 



75 



10 



11 



104 



The number of harmonics present in modal 
and falsetto register phonations 
produced by naive and singer subjects 
at three frequencies within the region 
of register over lap 80 

Example of an harmonic spectrum display 
for phonations produced in the modal 
and falsetto registers by one subject 
at 208 Hz 

Example of an harmonic spectrum display 
for phonations produced in the modal 
and falsetto registers by one subject 
at 262 Hz 

Example of an harmonic spectrum display for 
phonations produced in the modal and 
falsetto registers by one subject at 
330 Hz 3.06 



105 



zx 



CHAPTER I 



INTRODUCTION 



In recent years, a number of the parameters of 
human voice have been studied experimentally. For example, 
the acoustic waveform generated by the glottis has been 
approximated either directly or indirectly (Flanagan and 
Landgraf , 1968; Miller, 1959). Moreover, data have been 
accumulated about such laryngeal factors as 1) vibratory 
patterns (Timcke, von Leden , and Moore, 1958; Sonnesson, 
1960), 2) vocal fold length (Hollien and Moore, 1960, 
Hollien, 1960; Sonninen, 1954, 1962), and 3) vocal fold 
thickness (Hollien and Curtis, 1960; Hollien and Colton, 
1969; Hollien, Coleman, and Moore, 1968). Finally, sub- 
glottic air pressure and air flow have been studied by 
several experimenters utilizing a variety of research 
techniques (van den Berg, 1956; Ladefoged, 1962; Kunze, 
1964; Isshiki, 1964; McGlone, 1967, Murry, 1969). 

Although the quantity of such research is impres- 
sive, little effort has focused on the acoustic, phys- 
iological, or perceptual correlates of the vocal registers. 



This paucity of information about registers is puzzling 
since most writers agree that registers are useful and 
observable phenomena of the voice. In view of the research 
conducted in many other phases of the voice, it is appro- 
priate and essential for further investigation of vocal 
registers. In this regard, it is first necessary to define 
the term register and achieve a consensus of opinion about 
register correlates. 



Re gister definitions 

Several writers in such diverse disciplines as vocal 
music, voice science, laryngeal physiology, etc., have at- 
tempted to provide definitions of register. In spite of 
the different theoretical or practical approaches represent- 
ed by these varied disciplines, the definitions offered are 
remarkably similar to one another. 

For example, some writers in vocal music cite a 

definition similar to that offered by Manuel Garcia. In 

this definition, a register is: 

. a series of consecutive and homo- 
geneous sounds arising from the grave 
to the acute produced by the develop- 
ment of the same mechanical principle, 
the nature of which essentially differs 
from any other series of sounds equally 
consecutive and homogeneous produced 
by another mechanical principle. 

(from Mills, 1913) 



Mills (1913), Seiler(l868) , Vennard (1967, 196S). 



Two representatives from the medical field have 
also offered a d e f i n i t i o n : 

Registers are a series of consecutive 
voice tones of e q u a 1 t i in b r e . A t a 
certain point, the musical ear can 
distinguish a tone from another series 
of tones. The acoustical similarity 
among the tones of one series is due 
to a constant distribution of higher 
harmonics 

(Luchsinger and Arnold, 1965) 

The above definition is very similar to one offer- 
ed by two experimental phoneticians. Hollien and Michel 
(1968) state that "A register is defined as a series of 
consecutive frequencies of similar quality ..." but they 
also add. the qualification that ". . .in addition there 
should be little or no overlap between adjacent registers 
. . ." As may be noted, the latter statement is not found, 
in the two definitions previously quoted. Although of seme 
immediate interest, a detailed discussion of this specific 
criterion for a register will be postponed for a later 
sec t i on- -one in which frequency limits are considered in 
d e t a i 1 . 

In summary, a register may be thought of as a series 
of consecutive frequencies perceived to be similar in 
quality. At the same time, it is also possible that some 
adjacent registers share a small or even moderate range 
of overlapping, i.e., identical, frequencies. Hence, 



although no precise definition is available, those pre- 
viously quoted may serve as first approximations. 



Register terminolog y 

Although a number of authors from diverse fields 
of study are able to agree roughly ona definition of reg- 
isters, few partition this parameter of the human voice in 
the same manner. Con sequent ly--depending on its source--a 
description may include from one to five separate and dis- 
tinct registers. On the one hand, Seiler (1868)., Mills 
(1913), Hollien and Michel (1968), and others have postulat- 
ed three major registers, whereas Vennard (1967) and Luchsinger 
and Arnold (1965) have presented the' concept of only two 
registers. Vennard (1967), indeed, has argued that the 
"ideal" voice would exhibit only one register since, he 
states, it should be possible to vary frequency over 
a wide range with little or no noticeable change in other 
parameters of the fundamental laryngeal tone. However, 
Vennard concludes that only in a few individuals would such 
an "ideal" voice be found. At the other extreme, Horner, 
Fransson , an d Fant (1964) and Hoole (1902) have postulated 
five registers. However, even though authorities cannot 
agree on the exact number, the trend is to recognize three 
major registers with the possibility of one or two lesser 



ones located above or below the frequency ranges occupied 
by the major registers. 

Since most writers in this area fail to agree on 
exact register divisions, it is not surprising that a 
substantial number of different terms also have accumulated 
in the literature. In fact, Morner, Fransson, and Fant 
(1964) list 10 7 names that have been used in such defini- 
tions. While it is not the purpose of this paper to dis- 
cuss these terms in detail, a brief review is undertaken 
in order to illustrate labels currently in use. 

It is convenient to locate a register somewhere 
on a frequency continuum which encompasses almost all of 
the frequencies produced by the human voice. Consequently, 
the lowest register may be called vocal fry (Hollien and 
Michel, 196S; Hollien, Moore, Wendahl, and Michel, 1966; 
Wendahl , Moore, and Hollien, 1963), strohlbass (van den Berg, 
1960; Preissler, 1939), or deepest range (Morner, Fransson, 
and Fant, 1964). Proceeding up the frequency scale, the 
second register may be labelled chest (van den Berg, 1960; 
Smith, 1957; Sonninen, 1962; van den Berg, 1956, and others) 



or modal 2 (Hollien, 1960; Hollien and Michel,- 1968; McG 



cuione 



^Modal is assumed to encompass both the chest and 
mid registers. Hollien (1968) emphasized that the modal 
register refers to the normal speaking and singing range of 
the human voice, i.e., it is the phonatory range or register 
used most of the time by a speaker or singer. 



and B r own, 1 9 6 8 ) o r d e c p level ( M orne r , Fra n sson, a n d F ant, 
1964). Some authors postulate mid register (Luchsinger 
and Arnold, 1965; Large, 1968) or mid level (Morner, 
Fransson, and Fant , 1964). In addition, two authors uti- 
lize the terra falsetto I to refer to this register (Chiba 
and Fujiyama, 1958). Above the mid register may be head- 
voice (Rubin and Hirt, 19 60), falsetto II (Chiba and 
Kajiyarna, 195S) or high level (Morner, Fransson, and Fant, 
1964). However, most writers utilize the term falsetto to 
refer to this r egi s t er - --whi ch is located above the modal or 
mid register (MoGlone and Brown, 1968; Hollien and Michel, 
1968). Toward the upper end of the frequency scale, there 
may be a 1) whistle register (Preissler, 1939; van den Berg, 
1960), 2) flute (van den Berg, 1960) or 3) falsetto register 
(Luchsinger, 1953), or it may be called the highest level 
(Morner, Fransson, and Fant, 1964). 

Thus, although both the divisions of the voice and 
the terminology used to divide them are vague, it is possible 
that some meaningful acoustic, physiological, and/or per- 
ceptual parameters can be associated with each of the known 
registers. If so, perhaps they can be redefined with ref- 
erence t o t h e s e p a r a m e t e r s . 
Acou stic correl a t e s 

Previously, it was indicated that most writers located 



registers by referring to some portion of the frequency 
continuum. For example, Hoole (1902) suggests that the 
extent of the "lower register" in the tenor singing voice 
is from C^ (261 Hz) to B 4 (493 Hz). Hoole' s "head register" 
occupies a range from D 5 (587 Hz) to C g (1108 Hz). Hoole 
does not report these limits as data but apparently based 
them on experience and informal observation. 

Sonninen (1962) provides the following frequency 
limits for his one female singer: chest register, 139 to 
311 Hz; middle register, 311 to 622 Hz; head voice, 622 to 
1244 Hz. Hollien and Michel (1968) have reported the pho- 
national ranges of 23 subjects (12 males and 11 females) 
who phonated in the vocal fry, modal, and falsetto registers 
For vocal fry, the mean range reported was 24 to 52 Hz; in 
modal register the mean range was 94 to 28 7 Hz whereas in 
falsetto the mean low frequency limit was 156 Hz and the 
mean high frequency limit was 795 Hz. The mean frequency 
range of females in vocal fry was 18 to 46 Hz; for the 
modal register the range extended from 144 to 538 Hz, arri , 
in falsetto, the female group could phonate within a mean 
range extending from 49 5 to 1131 Hz. Thus, these three 
registers appear to occupy a specific range of frequencies 
although the exact frequency limits vary within and across 
s exe s . 



The data presented by Hollien and Michel (1968) 
also provide some evidence from which one could argue 
that registers are not divisible solely by their position 
on the frequency continuum. For both males and females, 
the" frequency limits of the modal and falsetto registers 
overlap to some degree. For example, the mean overlap be- 
tween these two registers for males is 131 Hz; for females 
it is 43 Hz. Although these ranges are small, nevertheless, 
it is suggested that some subjects can produce identical 
frequencies in the modal and falsetto registers. McGlone 
and Brown (1968) also present data in which some subjects 
displayed an overlap between the two registers. This evi- 
dence would suggest that frequency discontinuity, per se, 
is not the only means for distinguishing registers. Per- 
haps some other parameter, physiological, perceptual, or 
acoustic, is also required for differentiating registers. 

By definition, registers would also seem to be tone 
quality phenomena. Insofar as quality and wave spectrum 
are related, one would expect consistent spectral charac- 
teristics for each register in addition to or in place of 
the frequency differences which also exist. For. the chest 
register, Smith (1957) suggests that the "timbre [of the 
tone] is bright, due to the multitude of overtones." How- 
ever, he used rubber models of the larynx for his observations, 



and his conclusions may not accurately reflect the actual 
operation of the human glottis. Furthermore, Smith does 
not report the data upon which he based his conclusions. 
Sonninen (1962) reported similar data on one subject; he 
concluded that the number of partials in the chest register 
(frequency limits of 139 to 311 Hz), was 30. Sonninen does 
not report the exact fundamental frequencies at which the 
phonations used in his analysis were made nor does he spec- 
ify other parameters of the phonations which could possibly 
affect the distribution of the harmonic components (e.g. , 
vocal intensity or vowel). Finally the measurements were 
derived from Sonagrams which do not necessarily yield good 
frequency resolution. Also, Luchsinger and Arnold (1965) 
claim that registers can be differentiated primarily on 
the basis of the spectral components of the tone, but they 
do not report supporting data either. 

Large (1968) has reported the results of some work 
in which he compared the harmonic spectrum of tones produced 
in the chest and mid registers by female singers. He re- 
ports that when the subjects phonated at the same frequency 
and intensity level, the tone produced in the chest register 
displayed more energy at the higher harmonics than when the 
tone was produced at the same frequency/intensity level but 
in the mid register. Large ' s conclusions are based on a 



10 



Sonagraph analysis and the pictorial displays derived from 
such an analysis. However, it is known that the Sonagraph 
is not well suited to yield complete amplitude by frequency 
analysis primarily because of its limited dynamic range 
(35 dB). Furthermore, it would be premature to conclude 
that two registers exhibit different harmonic distributions 
when little is known about the expected distributions within 
any one register. Since it is possible that spectral char- 
acteristics of a phonation will vary as a function of 1) time, 
2) subject, and 3) vowel. In addition to the fundamental 
frequency and intensity, more normative data about the har- 
monic distributions within each register are needed before 
generalizations of this type can be made. 



Perceptual correlates 

In the absence of supporting data, a working hypoth- 
esis that registers manifest different spectral distribu- 
tions appears reasonable. Traditionally, it has been 
postulated that the perceptual correlate of wave spectrum 
is quality. Accordingly, it is interesting that in few, 
if any, of the experimental studies of the acoustic or 
perceptual characteristics of registers, has there been 
any attempt to relate the perceptual judgments (of tone 
quality) to the acoustic features of the judged phonation. 



11 



In addition, few procedures have been reported by which 
the reliability and validity of tone quality judgments 
could be evaluated. In fact, most authors apparently have 
utilized only informal evaluations as the means of deter- 
mining the register in which the tone was produced. As 
a consequence , there are few or.no data by which one 
can evaluate the consistency of register judgments. 

In some studies, the experimenter has utilized a 
voice "break" to differentiate one register from another 
(Rubin and Hirt, 1960). With this technique, the subject 
produces consecutive tones either ascending or descending 
in frequency. The voice "break" is located by observing 
the tone, or point between two tones, on which there is a 
change of quality. However, there is some evidence from 
which one might conclude that the register break, or shift, 
technique is not a reliable means for distinguishing reg- 
isters. McGlone and Brown (196S) investigated several 
methods by which one could determine the tone on which the 
register break occurred and reported that such could not 
consistently be identified by observers. In this respect, 
it would appear that the "break" technique may not be a 
reliable procedure to utilize in categorizing registers. 
Consequently, in any procedure for identifying registers, 
the investigator is forced to rely on the observer's 



12 



judgments of the differences in voice quality of two tones 
produced in two registers. 

Although tone quality differences may be the pri- 
mary cue to the differentiation of registers, it is not 
necessarily the only cue. For example, the frequency level 
of the phonation may be a feature used by a listener; more- 
over, loudness also may be a perceptual correlate. Vennard 
(1968) has commented that a falsetto register phonation is 
not as loud as a phonation produced in the modal register. 
Thus, the loudness of the phonation may contribute to the 
identification of a register. Nevertheless, the primary 
characteristic of a register appears to be the perceptual 
attribute related to the distribution of the frequencies in 
the complex tone. 



Ph ysiological correlates 

Assuming that the acoustic and perceptual correlates 
of each register could be established, then one could define 
the physiological mechanisms which underlie the production 
of registers. Although fragmentary and incomplete, some 
appropriate physiological data have. been reported already. 

Several experimenters have conducted investigations 
of the myoelastic properties of the larynx when tones are 
produced in different registers. Sonninen (1962) measured 



13 



the vocal fold length of one female singer when tones were 
produced in the chest, middle, and head registers. He 
found dif-ferent patterns of vocal fold length in the three 
different registers. Sonninen claims that the "lengthening 
of the vocal cords is basically linked to the question of 
registers." (See also Sonninen, 1954.) Hollien and Moore 
(1960) also observed that vocal fold lengthening patterns 
varied when their subjects produced tones in different 
registers. In fact, they reported three different patterns 
for the falsetto register alone. 

Hollien and. Col ton (1969) present some data from 
which one could conclude that vocal fold thickness in the 
falsetto register does not vary in the same manner as it 
does in the modal register. That is, vocal fold thickness 
decreases as frequency is increased in the modal register, 
whereas in the falsetto register, vocal fold thickness does 
not show a similar trend. 

In an o t h e r area of investigat i o n , T i m eke, von Leden, 
and Moore (1958.) have suggested that the vocal vibratory 
pat t e r n s (as s u m m a r i z e d i n s u c h t e r m s a s Open i n g Q u otisnt 
and Speed Quotient) show differences when a. subject phonates 
in the head or chest register. However, in this study (as 
in many of others discussed) the data have been 1) based, on 



14 



extremely limited subject samples and 2) confounded with 
those physiological processes possibly resulting from 
frequency differences between the two tones. Therefore, 
it is difficult to specify those patterns which are asso- 
ciated with the register and those associated with frequen 
chan ge . 



cy 



Baisler (1950) investigated glottal length, width, 
and area when phonations were produced in two registers. 
Two subjects phonated at identical or nearly identical 
frequencies in the normal and falsetto registers. In gen- 
eral, glottal length, width, and area decreased when a sub- 
ject changed from normal to falsetto. Baisler also found 
differences in the duty cycle of the laryngeal generator 
when a subject produced tones in the different registers. 
On the other hand, he found the vibratory patterns associated 
with the two registers were not consistent for each of the 
two subjects. This finding suggests that a vibratory pat- 
tern characteristic of o n e subject may not be characteristic 
of another subject, even when both subjects produced the 
tone at the same fundamental frequency and register. 

I ti addition to possible changes of myoelastic activ- 
ity as a function of register, certain aerodynamic activities 
may also show differences of magnitude and/or timing when 



15 



register is altered. For example, Isshiki (1964) has illus- 
trated the differences of air flow rate through the glottis 
as a function of register. Kunze (1962) has found that the 
intratracheal air pressure differed among vowels phonated 
in the modal register but these same vowels produced in the 
falsetto register did not show such air pressure variations. 
Finally, van den Berg (1956) failed to find any differences 
of subglottic air pressure (as measured by a catheter in 
the glottis) when register was changed. Murry (1969) found 
high intratracheal pressures for phonations produced in vocal 
fry, although his air flow data support McGlone's (1967) 
findings of very low air flows in vocal fry phonations. 

In summary, most of the acoustic, physiological, and 
perceptual characteristics of registers are based simply on 
first order observations and/or limited data. Nevertheless, 
it can be concluded that registers are indeed voice quality 
phenomena and are as being closely linked to the frequency 
continuum. Furthermore, it appears that registers are the 
result of certain quantifiable changes in the myoelastic 
and/or aerodynamic properties of the larynx. However, it 
should also be recognized that little research has been 
conducted wherein the focus of the investigation was 
on toward quantifying the differences of the glottal 



16 



tone spectrum which might be related to register differ- 
ences. On the other hand, it is obvious that a great deal 
of research has been conducted on glottal activity as a 
function of frequency and/or intensity change, but within 
an individual register. 3 Consequently, in many instances 
the acoustic and physiologic differences at the glottis-- 
thought to be associated with a r egi s t er --may also reflect 
the changes which, occur in response to frequency alteration. 
Therefore, in any study of registers, it would seem nec- 
essary to account for these glottal patterns resulting from 
register change and those which result only from frequency 
change. However, since registers tend to be associated 
with a range of frequencies, it is not possible to simply 
choose any frequency level in order to secure phonations 
from two registers. 

Fortunately, several authors (Hoi 11 en and Michel. 
1968; McGlone and Brown, 1968) have reported that some 
subjects can produce a range of frequencies in both modal 
and falsetto registers. Large (1963) has utilized an over- 
lap b e t w e 6 n t h e f e m a I e c h est and m i d registers fro m w h i c h 



Daaste (1968), Hollien (1950), Hollien, Coleman, 
and Moore (1968), Hollien and Michel (1968), Hollien and 
Colton (1969), Ishikki (1964), Michel and Hollien (1.968), 
Murry (1969), van den Berg (1956), van den Berg (1960). 



17 



to obtain his "isoparametric" tones. The overlap between 
the modal and falsetto registers is probably a more fruit- 
ful aire a to secure phonations which differ only in register, 
since the modal/falsetto distinction is more likely to be 
recognized by trained and untrained subjects. In this 
way perceptual judgments of tone quality differences could 
be obtained. 

This investigation was designed to obtain data on 
the acoustic and perceptual correlates of the modal and 
falsetto registers, at least for the region where these two 
registers overlap along the frequency continuum. It is also 
of interest to determine how stable these correlates are 
for 1 ) the same subjects over a period of time, 2 ) dif- 
ferent subjects, and 3) subjects who possess extensive 
training and/or experience in singing. Specifically, the 
focus of the study was directed toward the following ques- 
tions: 



1 . W hat is the range of frequency overlap bet w e e n 
the modal and falsetto registers for trained 
singers and untrained subjects? Do singers 
differ from, non- singers on this characteristic? 

2 . Ho w C onsistently are s u bjects able to prod u c e 
t h e r e g i s t e r o v e r 1 a p? 

3 . W h a t a r e t h e vocal in t e n s 1 1 y 1 i m i t s for the 
two registers within the range of register 
over! ap? 



Are naive and trained observers able to 
discriminate phonations produced at the 
me fundamental frequency and intensity 
in different registers? 



s 
but 



If observers are able to discriminate 
between phonations produced in different 
registers, are they able to place the 
phonations into two distinct categories, 
utilizing a voice quality criterion? 



How are the harmonic components of 
phonations produced in the two registers 
distributed? Does this acoustic feature 
change when subjects produce phonations 
at different frequency levels within the 
region of register overlap? 



CHAPTER II 
PROCEDURE 



T he major focus of this investigation was to 
study some perceptual and acoustic correlates of prona- 
tions produced at the same frequency and intensity levels 
but in different registers, i.e., modal and falsetto. 
First, the range of frequencies an individual could pro- 
duce in each register was determined for a large group of 
adult males selected at random from the University of 
Florida student, population. From this larger group, two 
subgroups were identified: these were 1) naive or un- 
trained subjects and 2) singers. Second, vocal intensity 
ranges at three selected frequencies within the register 
overlap were obtained for eight individuals in each of 
the two subgroups. Third, the 10 subjects (five from 
each group) who exhibited the most homogeneous phon a t i onal 
ranges in each register were selected to produce phona- 
f i o n s at t h rec s p ecifie d f r e q u e r. c i e s w i. t h in t h e registe r 
o v e r 1 a p . T ape r e c o r d i n g s o f ' t h e s e p h o n a. t i o n s w e r e p 1 a. y e a. 
to naive and experienced observers who were asked to 
1) discriminate between samples produced in the two registers 



19 



20 



and 2) on the basis of voice quality, place each phona- 
tion into one of two categories. Finally, the pho nations 
were analyzed with a wave analyzer in order to determine 
the distribution of the harmonic partials contained in each 
of the signals. 

S u b ,1 e c t s 

Forty-five adult male volunteers (35 naive and eight 
trained singers), 18 to 25 years of age were elicited from 
the general student population at the University of Florida. 
No initial restriction was placed on the acceptance of these 
volunteers except that they did not exhibit a speech or 
voice abnormality. In addition, effort was made to ob- 
tain 1) individuals who had no training or experience in 
vocal music (hereafter called the naive group) 1 anc j 2 ) vol- 
unteers who possessed experience and/or training in singing 
(hereafter called the singer group). The singer group con- 
sisted of males with at least two years of active singing 
experience. All of these individuals had studied voice 
and had been enrolled in the University chorus or other 
singing groups in the University for "several years. 



It is understood that the term "naive" means un- 
trained in the vocal technique. 



21 



Ph o na tio n a 1 range procedu res 

P honation a l ran g e determinat ion. The ph on at ion a] 
range of each volunteer was determined for each register 
by requiring the subject to match the fundamental fre- 
quency (hereafter f ) of his production of the vowel /a/ 
to sinusoidal frequencies produced by an audio-oscillator 
(Hewlett-Packard 20 1C). These reference frequencies were 
presented at the tone/semitone intervals of the equal 
tempered musical scale. At each interval, the subject and 
the experimenter decided whether or not the subject had 
phonated that frequency in the intended register; these 
decisions were especially critical within the range of 
register overlap. When the subject could produce identical 
frequencies in the two registers, he was asked to alternately 
phonate that tone in the two registers. If the two phona- 
tions sounded perceptually dissimilar to the subject and 
the experimenter, it was assumed that the phonation was 
produced in different registers. It was recognized that 
there were limitations to this procedure, but it appeared 
to be the best a v a i. 1 a b 1 e to d i s t i n g u i s h the t w o registers 
within this region of overlap. Finally, when the subject's 
lowest and highest frequency limit in each register was 
obtained, a tape recording was made. Later, these record- 



ings w e r e c o n v e r t ed to a v i s u a 1 t 



i' a c e v 



i a a ph on e 1 1 e gr a ph 



22 



The resultant phonel 1 egr ams were measured in order to 
determine the actual frequency limits of each register 
for each subject. 

Phonational range stability . One of the purposes 
of this investigation was to determine the stability of the 
upper and lower frequency limits of the range of overlap 
between the two registers (modal and falsetto). There- 
fore, each of the subjects reproduced his phonational range 
each time he was seen. Reported in this study are the re- 
sults of the 10 subjects whose phonational ranges were 
determined on three separate occasions. These meetings 
took place approximately two and three weeks following the 
initial session. In order to control any possible time 
or day effect upon the data, all subjects returned at ap- 
proximately the same time during the day. 

Vocal intensity procedures 

The 16 volunteers (eight singers and eight naive 
subjects) who exhibited the largest and most homogeneous 
ranges of frequency overlap between the two registers were 
asked to return so that an estimate of their vocal intensity 
ranges could be obtained at the three frequency levels with- 
in the range of register overlap. These frequency levels 
represented, to the nearest semitone, the 25, 50, and 75th 



23 



percentage points of the overlap itself. At these three 
points, the vocal intensity ranges were determined by ask- 
ing the subject to produce the desired frequency at first, 
the most comfortable, then the minimum, and finally the 
maximum vocal intensity levels. In order to assist the sub- 
ject in producing the desired frequency, a reference tone 
was presented to which the subject matched the f of his 
own phonation. The subject's phonation was passed through 
a band-pass filter adjusted in such a manner that only 
the fundamental component of the tone remained. The 
frequency of the filtered signal was displayed on a counter 

(Hewlett-Packard 521A) and could be compared to the f of 

o 

the standard tone. 

All subjects were permitted to practice the produc- 
tion of the specified tone at the intended intensity level. 
When the subject phonated at that level, the rms sound 
pressure level (SPL) of the phonation was measured with 
the voltage meter of a Bruel and Kjaer microphone ampli- 
fier with its condenser microphone placed nine inches from 
the subject's lips. All readings were taken to the nearest 
.5 d.B with the meter set in its "slow" or heavily damped 
m ode; . M e a sure m e n t s w ere m a d e in r e f e r e nee to on e mil 1 i v o J t . 

Upon obtaining these intensity ranges, five singers 
and five naive subjects were selected who could produce 



24 



the most homogeneous phonational range in the two registers 



Perceptual procedur es 

Another major purpose of this investigation was to 
ascertain how well observers can distinguish phonations 
produced in the two registers when the fundamental fre- 
quency and intensity of the phonations are identical. 

Stimul i . The final task for all subjects was to 
produce a two or more second sample of the vowel / Q / at the 
three frequency levels within the region of frequency over- 
lap. As previously stated, these frequency levels cor- 
responded to the 25, 50, and 75th percentage points along 
the range of overlap. However, since all 10 subjects pos- 
sessed very similar phonational ranges in the two registers, 
the actual frequencies corresponded to 20S , 262, and 330 Hz 
for all subjects. These frequencies were chosen in such a 
way that the highest frequency (330 Hz) was at least two 
semitones from the upper limit of the modal register, and 
the lower frequency (208 Hz) was two semitones above the 
low falsetto limit. Thus, although they did not exactly 
correspond to the desired percentage' points, the frequencies 
utilized were very close to these levels. 

Each subject produced the specified frequency at 
his own most comfortable intensity which he monitored by 



25 



viewing the deflection of a General Radio Sound Level 
meter. Five examples of the phonations produced in the 
two registers at the three frequency levels were secured 
although only the most stable phonation, judged by the 
experimenter on the basis of frequency and intensity, was 
used in the subsequent perceptual procedures. 

All experimental tapes were presented in a good 
acoustic environment using an Ampex 351 tape recorder, 
Marantz amplifier, and AR-3 speaker system, or an Ampex 
601 tape recorder and playback system. 

Observers . Tw o observer gro u p s were utilized in 
the discrimination and categorizations procedures. Ob- 
server Group I consisted of 15 college-age students drawn 
from the University population who possessed no formal 
training in speech, voice, or singing. The other group, 
Observer Group II, was subdivided according to the ob- 
server's experience. In this way, the influence of the 
observer's training and/or experience was evaluated. The 
first subgroup consisted of singers or voice students who 
possessed extensive experience in singing or who had re- 
ceived formal training in vocal techniques. The second 
subgroup was composed of experimental phoneticians or 
persons interested in normal voice qualities. The third 
subgroup consisted of speech pathologists and speech 



26 



therapists drawn from the University speech and hearing; 
clinic . 

R egister di scrimination proc edure . The p u r pose of 
this procedure was to determine if a phonation selected 
from one register was perceptually different from a pho- 
nation selected from another register. At each of the three 
frequency levels, each subject's modal and falsetto pbo- 
nations were paired in an AX fashion and placed in a tape 
with the phonations from the other subjects. The same 
procedure was followed for each of the other frequency 
levels. Consequently, the frequency of the phonations 
presented within one experimental condition was identical. 
Thus, one subgroup of 20 phonations might consist of naive 
subjects phonating at 208 Hz whereas another subgroup of 
phonations might consist of singer subjects phonating at 
330 Hz. The 30 observers responded to each phonation pair- 
by circling S (for same) or D (for different). The pho- 
nations were one second in duration and separated by one- 
sec o n d. o f s i 1 e nee . 



oi 



Reg ister categorization procedure . Th e pu rpa 
this procedure was to determine if phonations produced in 
the two registers could be sorted into two voice quality- 
categories. Two procedures were devised to obtain this 
information; they are as f o 1 1 o w s . 



27 



1 • Force d c hoice procedu re . " T wo-sec o nd s a m pies 
of the pho nation, s produced at the same frequency by the 
10 subjects were arranged into random order onto an ex- 
perime n t a 1 t a p e . Cons e quently , there were three sets of 
20 phonations, each set consisting of phonations produced 
at a different fundamental frequency. The two groups of 
observers were asked to separate the 20 samples into two 
voice qualify categories. They were told that they must 
place 10 samples in one category and 10 samples in the 
other. Therefore, they knew that there were an equal 
number of phonations in each group. N training in or 
explanation about voice quality identification was given. 
However, in order to secure stable observer judgments, 
each tape was played three times; the first time so that 
the observers could decide which voices were representa- 
tive of the two categories, the second time to permit the 
responses to be recorded, and a third time so that the 
r e s p o n S e s c o U Id be verified by the obser v ers. 

2- F ree choice procedure . in this procedure, all 

phonations reg a i • d 1 e s s of f r e q u e n c y a p p ear e a i n d i v i ci u all y 
in one tape. Thirteen naive observers listened to each 



2 

This procedure is very similar to the procedure 

used by Michel (1964) to separate vocal fry phonations 
from harsh voices. 



28 



phonation and placed it into one of two voice quality 
groups. The characteristics of these two categories were 
established by the observers. There was no restriction on 
the number of labels to be used nor was any training in 
voice quality . gi ven . However, prior to presenting the 
experimental samples, 30 experimental samples were played 
so that the observers could practice the task and establish 
his criteria for the subsequent rating of the samples. 

Observer reliability . It was of interest to de- 
termine how consistently the observers placed the phona- 
tions into the two voice quality groups. Specifically, 
the question posed was if an observer placed a phonation 
into the same category on one occasion, would he place that 
same phonation into the same category the second time he 
heard it? The measure of within observer reliability was 
estimated by presenting the identical phonation twice in 
the course of the experimental sessions to 14 experienced 
observers and 10 naive observers. These same observers 
took part in the other forced choice and paired comparison 
perceptual procedures. In the free choice procedure, the 
13 observers heard the phonations twice also. The two re- 
sponses to the same phonations were compared and as a mea- 
sure of within observer reliability, a phi coefficient 
was c om put ed . 



29 



Acoustic measurements 

The final aim of this project was to investigate 
the possible spectral differences between the modal and 
falsetto registers. A sample approximately two seconds 
long and identical to that used in the perceptual pro- 
cedures was made into a tape loop and played at a constant 
tape transport speed (15 ips) to provide the input for a 
General Radio 1900A wave analyzer. The center frequency 
of the 50 Hz analyzing band was begun at Hz and increased 
to about 8 kHz. This upper limit was chosen as preliminary 
analysis of these samples revealed no significant energy 
peaks (defined as at least 2 dB above the voice level) above 
8 kHz. A line spectrum display was prepared from these 
amplitude by frequency records and the number of partials 
measured in each phonation was computed. The number of 
partials comprised the criterion measure for the subse- 
quent statistical tests. 



CHAPTER III 



RESULTS AND DISCUSSION 



Three major procedures were utilized to study several 
acoustic and perceptual correlates of the modal and falsetto 
registers. In the first procedure, estimates of the total 
phonational range of the two registers were obtained from a 
large group of adult males (N = 43). Subsequently, two sub- 
groups of eight individuals were selected from this larger 
population (the singer and naive subject groups). The range 
of vocal intensity at each of three frequency levels chosen 
as percentages of the range of register overlap was measured. 
In another procedure, phonations produced in the two registers 
by five singers and five naive subjects were presented to 
naive and experienced observers who were asked to differentiate 
these stimuli on the basis of voice quality. In the final 
procedure, the harmonic spectrum of the tone produced at each 
of the three frequency levels in the two registers was analyzed 
Since the 10 experimental subjects were screened for their 
phonational range in each register every time they were seen, it 
was also possible to obtain consistency estimates of subject 
phonational range. 



30 



31 



Phonatio n al__, T&ti£Bs_in_tJne_ moAal_and_^als e tto registe rs 

Table 1 presents the mean upper and lower semitone 
limits of the modal and falsetto registers for the 35 naive 
subjects and the eight singers. The average low modal limit 
of the naive subject group is slightly greater than 30 semi- 
tones FL whereas the upper limit of this register is about 
49 semitones. Singers, on the other hand, exhibited a 
slightly lower low modal register boundary and an upper modal 
boundary which was three semitones greater. The low falsetto 
limit for the naive subjects was 46 semitones FL whereas 
singers exhibited a low falsetto limit which was two semitones 
lower than the naive group. T-tests were performed on the 
differences between the semitone limits of each register for 
the two subject groups at each session 1 and are shown in 
Table 2. No significant differences were obtained between 
the two subject groups at any register limit in any screening 
session. Furthermore, t-tests computed on the differences 
between the mean total phonational range of the two groups 



The first session correspo n cl e cl t o that at w h i c h t h e 
3 5 naive and eight singers were screened. The second session 
w a s the s e s s i o n a t w h ich t h e v o c a 1 intensity d a t a w ere col 1 e c t e d 
In the third session, five subjects in each group produced the 
p h o n a t i o n s u s e d in the perc e p t u a 1 p r o c e d u r e s . 

2 
Total phonational range is herein defined as the dif- 
ference in semitones between the upper falsetto register limit 
and the lowest modal register limit. 



32 



Table 1. Mean phonational range of singers and naive sub- 
jects in the modal and falsetto registers. All values are 
in s e m itones re. 16.35 H z . 



Register Li m i t 



Total Modal 
Register Range 

Low Modal Limit 

High Modal Limit 



Modal Standard Deviation 

Naive* Singers** Naive Singers 



18.5 22.7 
30.4 29 . 5 
48 . 9 52.2 



2.9 
2.9 



2. 4 
1 . 



Total Falsetto 
Register Range 



Low Falsetto L 



imi 



High Falsetto Limit 



17.8 21 . 4 

4 5.9 43.0 

63.7 64.4 



4 . 3 



3 . 1 



2 . 7 



3 . 3 



Total Phonational 
Range 



33.3 34.9 



*N = 3 5 
**N r 8 



3 3 



- 


Cd 




>> 


S3 


1 


u 





■H 


cd 


£ 


= 


T3 


P, 


CD 


a 




CO 


3 


H 




o 


cd 


£ 


.Q 


-p 
o 


•H 


U 


-p 


CO 







CD 


-P 





3 


CO 


A 


H 


•H 


-p 


cd 


CD 




> 





X! 




fc 


£* 


H 




cd 


H 


-d 




< 


o 


- 




cd 


CO 







u 






CD 


CO 


■p 


-p 


d 


cd 


CO 







•H 


•H 


to 


to 


CO 


4-> 


0) 


CO 


O 


fc 


CD 


cd 




CO 


T1 


o 




.Q 


-p 


CD 


3 


-p 


CD 


CO 


a) 


U 




CO 


A 


h 


H 


■P 


CD 


cd 




to 


«H 


-P 


d 




cd 


•H 


^3 




CO 


C 


•o 




cd 





•a 




d 


d 


i-H 


~H 


Cd 


cd 


cd 




•o 


-p 








X! 


> 


e 





■H 

Cti 


CD 


CO 


c 


-d 


cd 




-p 


5 


d 









d 


CD 





0) 


-P 


^ 





cd 








-p 


> 


s 


CD 


+-> 


■H 


X.' 


CD 


-P 




Si 


CO 


CO 




CD 





& 




o 


cd 


xi 


c 


M 


V 


a> 


H 


cd 


W 


CD 


f'-1 



•H 


CD 


o 




Q 


£ 


to 









c 






■P 


cd 




• 


•H 


Jh 


-t 


CM 


£ 




Em 




CD 


H 




CD 


CO 


cd 


CO 


i-H 




r* 


CD 


.O 








r* 


cd 


,d 


■H 





H 


+J 


-P 


■P 





O 






■H 






•P 




H 


cd 





cd 


S 


to 


-p 





d 


O 


gH 


cd 


Eh 


£ 


Oh 
D, 

cd 

H 


> 

o 










-p 




-p 




0) 


,d 


CO 


to 


H 


•H 


cd 


W 


Eh 









■p 




-p 









CO 


& 


H 





cd 


►H 


Ph 



cd 
O 

-1 s 



o 

■H 

CO 
CO 


CO 



t> 






\t< 



CM 



CM 
CM 



CM 

CM 






•* 



^ 



Si 


cd 


en 


^ 


to 


T3 


■ 


• 


•H 





CO 


CM 





d 




a 





Cj 


o 


■H 





H 


CO 


■H 


CO 


CO 


CO 


CO 





CO 





CO 





w 


'O 


w 


p 


c 


■o 


CO 





?H 


In 


V 


•H 


■H 





X! 




CO 


H 



-p 

CO 

-P 

•a 


H 
■H 

cd 

-p 




> 




CO 

o 




Si 
-p 

P 1 
ca 



cd 
o 



to 

■H 
CO 



3 4 



also failed to reach significance. It would appear, there- 
fore, that both the naive and singer subjects in this study 
possessed very similar phon at 1 onal ranges in the modal and 
falsetto registers as well as total phonational range. 

The finding that singers and naive subjects exhibit 
similar phonational ranges is somewhat unexpected since most 
authors believe that singers exhibit larger phonational 
ranges. For example, Luchsinger and Arnold (1965) expected 
that naive subjects would possess smaller phonational ranges. 
In this study, perhaps, the singers possessed a smaller pho- 
national range than would be normally expected. However, 
previous data provided by Preissler (1939) suggests that the 
phonational ranges for the singers in this study are not ex- 
cessively small. For example, Preissler reports total ranges 
of 25 to 54 semitones for his sample of 232 singers. Preissler 
does not report the mean phonational range for these subjects, 
but from his data one can estimate an approximate average 
range of 36 semitones. The mean phonational range of the 
singer subjects in this study was 35.1 semitones. It would 
appear t h a 1 t h e p h onatio n a 1 ranges of the singers in t h i s 
study were similar to those reported by Preissler. However, 
the results of this study and Preissler' s do not agree with 
the results reported by Nadoleczny (cited in Luchsinger and 
Arnold, 1965). Nadoleczny found phonational ranges of four 



35 

to five octaves, somewhat larger than the 3 to 3\ octave 
ranges of Preissler's subjects. However, Nadoleczny re- 
ports the data in terms of absolute range, not average 
ranges. Comparison of these ranges suggests good agree- 
ment between Nadoleczny (4-5 octaves) and Preissler (2-5 
octaves). In this study , the smallest singer phonational 
range was 28 semitones, the largest 41 semitones. In gen- 
eral, the results of the three studies are reasonably simi- 
lar. However, perhaps the ranges of the singers in this 
study are slightly smaller. 

The naive subjects in this study exhibited total pho- 
national ranges of about 33 semitones. This result agrees 
very well with the data reported by Hollien and Michel (1968) 
and Hollien, Dew, and Beatty (1969). The 12 naive male sub- 
jects of Hollien and Michel exhibited a mean phonational 
range of 33 semitones. 3 Hollien, Dew, and Beatty (1969) re- 
port a mean total phonational range for their 332 male sub- 
jects of 39 semitones. These data agree very well with the 
data from this study and suggest the conclusion that the 
total phonational range of naive subjects obtained in this 
study was not exceedingly large. 



Hollien and Michel (1968) report their data in tones 
FL. For purposes of comparison, their data has been con- 
verted to semitones FL . 



3 6 



The results shown in Table 1 reveal that, on the aver- 
age, naive subjects achieved a phonational range in the modal 
register of about IS. 5 semitones whereas in the falsetto 
register, they achieved a range of 17.8 semitones. On the 
other hand, Hollien and Michel (1968) recorded a mean modal 
register range of 19.3 semitones for their 12 adult males 
and, in the falsetto register, only a 14.4 semitone range. 
However, these latter data are based on a smaller subject 
sample (12) than the ones used in this study. Consequently, 
it is very possible that the mean frequency limits of the 
two registers for the naive subject group in this study may 
be more realistic estimates of the frequency limits of the 
two regi sters . 

As stated, Figure 1 is a plot of the data shown in 
Table 1. A E an additional index of the variability, "the absolute 
lowest and highest frequency measured at each register 
boundary is displayed. For both groups, there is a moderate 
range of frequencies about the mean limit for each register. 
But, in general, the average frequency reported in Table 1 
appears to bo a good indicator of the frequency limits of 
both registers for the population of naive and singer sub- 
j e c t s . 



37 




NA!V; 

(! 



UBv'EC 

35) 



MODAL FALSETTO 
S ! N G E R S 



F i g ii r e 1 . M e an ph o n a t ional ranges o f naive and singer 
s u b j e c t s i n 1. h e m o d a 1 a n d falsetto re g i sters . T h e e n d 
points of the solid bars indicate the mean upper and 
1 o w e r limits of each register . The dotted lines i n d i C a. t e 
the group ranges measured at each limit. 



38 



P h o n a t i o n a 1 range consistency 

Table 3 presents the three estimates of the phona- 
tional range for the experimental subjects. A comparison 
of the data presented in Table 1 and Table 3 suggests that 
these 10 carefully selected subjects exhibited a wider pho- 
national range (in both registers) than did the subjects in 
the larger and more heterogeneous subject group. 

The significance of these data, however, is that the 
phonational range in both registers measured over a period 
of time is remarkably stable. The largest standard deviation 
in this data is slightly greater than one semitone; the 
smallest is .2 semitone. Furthermore, there appears to be 
a. tendency for greater variability to occur at certain register 
limits. For example, the low modal register limit of the 
naive subjects is the most variable whereas the high modal 
limit is the least varaible. This trend may be the direct 
result of the procedure utilized to determine the phonational 
ranges of the subjects. The low modal limit was tested first, and. 
it is possible that the subjects did not have sufficient 
practice to achieve a high degree of consistency in attain- 
ing the low modal frequencies. Perhaps, if the order of 
test i n g w e r e v arie d , t h e t r e n d w o u 1 d h a v e b e e n d ifferent. 

Interestingly, the upper modal and lower falsetto 
exhibit little variability. One might expect greater 



39 



Table 3. Mean phonati onal ranges of tie modal and falsetto 
registers fornaive and singer subjects on each of three pho- 
national range screening sessions (M = 5 per group). All values in 
semi tones FL . 



Mean 
Naive Singers 



Standard Deviation 
Naive Singers 



First Screening; 



Low Modal 
High Modal 
Low Falsetto 
High Falsetto 



26 . 9 
53 . 2 
4 2.7 
66 . 9 



28 . 6 
53 . 5 
40 . 8 
67.1 



5 . 4 
.9 
. 7 

1 . 6 



6 . 2 
2.9 
2. 
1 . 3 



Second Screening 



Low Modal 
High Modal 
Low Falsetto 
High Falsetto 



29 . 3 
53 . 2 
41 . 4 
66 . 



29 . 4 
55 . 3 
41 . 2 
67 . 2 



1 . 2 

.9 

1 .3 

1 . 9 



1 .1 

3 . 3 

3 . 

1 . 5 



Third Screening 



Low Modal 
High Modal 
Low Falsetto 
High Falsetto 



26 .9 
53 . 5 
40 . 9 
66 . 



28 . 4 

55 . 4 

39 . 6 

66 . 5 



1 .6 

1 .6 

.9 

2.4 



1 .4 
2. 1 
3 . 7 
1 . 9 



Mean of the Threi 



Ses 



si on s 



Low Modal 
High Modal 
Low Falsetto 
High Falsetto 



27 . 7 

53 . 3 

41 . 7 

66 . 3 



28 . 8 
54 . 7 
40 . 6 
66 . 9 



1 .3 

. 2 
. 9 
. 5 



. 5 

1 . 




40 



variability since these limits were the most difficult to 
obtain. Apparently, using the screening procedure described 
in this study, one can obtain reliable estimates of the fre- 
quency limits of the modal and falsetto registers. It was 
also observed that individual phonational ranges were stable 
over time. Eor example, most individuals varied no more than 
two semitones from one session to another. 

Little other research has been reported in which the 
variability of register frequency limits has been discussed. 
Hollien, Dew, and Beatty (1969) report that 33 of their sub- 
jects produced a "total" (i.e., modal plus falsetto) phonational 
range on at least two occasions. They found that the varia- 
tion from one occasion to the next usually was no greater 
than one or two semitones. McGlone and Brown (1968) secured 
three repetitions of their subjects producing tones on which 
there was a register break. At the upper modal limit, they 
report an average difference (not standard deviation) from 
one repetition to another of 2.1 semitones. At the low falsetto 
limit, the mean difference was one semitone. In this study, 
the mean difference from one condition to another at the 
high modal register limit is about 1.2 semitones whereas at 
the low falsetto limit, the average difference is .7 semi- 
tones. In spite of differences of sample size, the results 
of the two studies are remarkably similar. Since McGlone and 



41 



Brown relied on the register "shift" as a definition of 
register limits, one might expect even greater differences 
of the upper modal and lower falsetto register limits between 
the two studies. 

One feature of the phonational ranges of ihe modal 
and falsetto registers of particular interest in this study 
was the number of frequencies a subject could produce in both 
registers, or as previously defined, the register overlap. 
From the data in Table 1, one can compute a three semitone 
overlap between the two registers for the non-singers and a 
nine semitone overlap for the singers. The overlap in this 
study is larger than that which was found by Hollien and 
Michel (1968) (i,e., .7 semitone). Using the register shift 
procedure previously discussed, McGlone and Brown (1968) de- 
fined the upper limits of the modal register and the lower 

4 
limits of the falsetto register. From their data, a 2.6 

semitone overlap was computed. The three semitone naive 

subject overlap reported in this study is slightly larger 

than the overlap computed, from McGlone and Brown's data and 

much larger than that computed from Hollien and Michel's data, 

Since the number of subjects used in this study ( N = 35) was 

much larger than either of the other studies, the three 



The authors use the term "upper register" which ac- 
cording to Brown (1969) refers to the "falsetto register. 



42 



semitone overlap may be a more precise estimate of the actual 
overlap. Furthermore, meticulous care was exercised to obtain 
the largest possible overlap for each subject. 

Further, it can be noted in Table 3 that, for all 
three sessions, the singers exhibited a larger amount of over- 
lap than did the naive subjects. Statistically significant 
differences were found between the register overlap of 
the naive and singer groups (N = 35). The overlap between 
naive and singer subjects obtained in the second and third 
sessions was not statistically different; suggesting that 
these subjects were much more homogeneous than thelarger 
group. It would seem therefore that the five subjects in 
each group who produced the phonations used in the perceptual 
studies exhibited very similar ranges of register overlap. 

In summary, it was found that naive subjects exhibit, 
on the average, a modal register phonational range of 18.5 
semitones and a falsetto range of 17.8 semitones. Singers 
exhibit slightly larger ranges than, but not significantly 
different from, naive subjects, i.e., 22.7 semitone modal 
range and 21.4 semitone falsetto range. The naive subjects 
studied produced a three semitone overlap between the modal and 
falsetto registers whereas singers exhibited a nine semitone 
overlap. Finally, repeated estimates of phonational ranges 
of singers and naive subjects are very similar to each other 
which indicates the stability of a. phonational range estimate. 



4 3 



r ocal in ten s J t y 1 jj i t . 



Each subject produced the minimum, most comfortable, 
and maximum sustainable intensity levels at each of three 
frequencies within the range of register overlap. The re- 
sults of this procedure are presented in Table 4. More- 
over, Figure 2 is a plot of the vocal intensities at each 
level for the naive subjects. Figure 3 is a similar plot 
for the singer subjects. In general, the magnitude of vocal 
intensity increases at each level of intensity for pronations 
produced in the two registers. However, vocal intensity does 
not increase in the same way in both registers. For example, 
the minimum intensities produced in the two registers axe very 
similar. But, at the most comfortable intensity condition, 
the intensities produced by the naive subjects in the modal 
register are about six dB greater than the intensity level 
produced in falsetto, and singers achieve about 1.1 dB dif- 
ference between the two registers. At the maximum intensity 
condition, the intensity difference between the two registers 
is much more marked. In general, it is evident that the 
intensities of falsetto register pronations are much lower 
than modal register pronations. In this regard, Vennard 
(1967, 1968) has observed that pronations produced in the 
modal register are louder t tan those produced in the falsetto 
register pronations. The vocal intensity data suggests that 



44 










> 






•H 


<H 




Cd 


o 




a 

u 










'H 




<p 


be 





01 


^t 




m 






0) 







p 


X! 




M 


P 


p 


■H 




H 


bD 


£ 








■H 


> 


S-i 


XI 


■H 




4-> 


H 


O 


•H 


H 


-P 


£ 


■H 


P 




s 


(1) 


co 




DQ 


rH 


H 


H 







cd 


> 




<H 










H 


U 


■o 






c 


>> 


m 


cd 


o 


■o 


H 





g 


Hi 


2 


•H 


•ti 


tr 




O 








s 


Sn 


h 




<H 


Cd 


0! 






A 








-P 





H 




U 


-Q 


C 


xi 


cd 


■H 


-p 


-p 


en 


-p 





H 


cd 


Xi 


S) 




P 


> 


to 







c 


C 


H 


■H 

P 


•H 


>> 


«J 


in 


P 


a 





•H 





3 


cn 


X! 


H 


fl 


a 


cd 



4-> 


CD 


> 


C 


P 




■H 












£V 


H 


■n 


cd 


Cti 


.n 


rH 


O 


^j 


!h 



O !D 

> > 





bD 



cd b 
i-f cd 



CD 











bfl 





£ 


p 


■H 


-p 


CQ 







ID 




H 




td 





h 


> 




•H 




td 




z 





ho 
C 

•H 

rH CO 

a 
•a 
o 

S 
> 

•H 



p 




•p 




cn 




^ j 


H 









u 

-X 

"^J I j 

O" CD 

> 

U 

Ph >-h 



CO 


cd 


00 


o 


£- 


CD 


^f 


r> 


to 


C- 


CO 


H 



to 

CO 



^ 



00 
CD 



CD 

CO 



CO 05 t> 
H f 00 



O CD M 



t> 05 H 
00 OS 00 



cn o co 
O <* oo 

CO Oi N 



J3 

cd 
p 
U 

o 



p 

CD 




CO 

CM 




hO 



cd 
f4 



05 



t~ 



05 

CD 






CN 



O 



•H 

d 

■H 



O Ol CO 



CO CM H 



CO 03 CO 

O CO rH 



<tf 



CM 

05 



CO O CO 



^ rH O 
00 O CO 



X! 

cd 

p 



y cd 




bD 
fi 
cd 



O 
CO 



05 

CD 



35 

t> 



H 



O 05 o 



rH G5 


CO 


05 


rH 


CO 05 


00 CO 


H 


CD 


00 


00 H 



co co m 
o t> r- 

C50 CO ^ 



oo co m 

•sT 1 CO ^ 

05 o cm 



^ CD CO 

CO t> co 

CO 05 Cd 



£3 

cd 
p 

o 



CO 

ts 



o 

p 





A 
p 

fi 


& 
p 

£1 

Ph 

cd 

H 

u 



> 

C 

>» 

CJ 

a 



cr 


<H 

<P 

o 


b.O 
C 

cd 

Sh 


Xi 

-P 

<H 

o 







o 









p 

CD 
•H 

bD 

U 



45 




1 i 

2 5 50 75 2 5 50 75 

r rvu.vWti^ui i-\-L c a r ri c z> o c u 
PERCENTAGE OF RANGE 
REGISTER OVERLAP 



iin 



Figure 2. Mean vocal intensity of naive subjects 
at the frequency levels of the -range of register 
ov e r 1 a p i n the rn o d a 1 and falsetto registers 



Modal 



Legen d 
Intensity Condition 



F a 1 s e 1 1 o 



© — . © 
A— A 



Mi n imum 

Most Comfortable 

Max! muni 



o—o 

A— —-A 

n — -o 



05- 



100- 



9 5- 



0" 



o rr 



r~ GO- 



cd 75 ~ 
-a 



70 



65' 



60- 



* 



L 



o 



o 



A 



O 



46 



D--. 



S 



/ 



-u 



D 



/ 



-A 



/ 



• 



y 



■o 



25 50 75 25 50 75 

FREQUENCY LEVEL EXPRESSED IN 
PERCENTAGE OF RANGE OF 
REGISTER OVERLAP 

Figure 3. Mean vocal intensity of singers at each 
frequency level in the modal and falsetto registers. 



Modal 



Legend 
Intensity Co n d i tio n 



F a 1 s e 1 1 o 



© — -Q 

A A 

m — n 



Hin i mum 

Most Co m fortable 

M a x i m u m 



O—O 

A—A 
□ — D 



47 

since falsetto register phonations are produced with less 
vocal intensity than are modal register phonations, one 
would expect that perceptually they would differ in loud- 
ness. Of course, it is not known if the lower vocal in- 
tensities of falsetto register phonations are the necessary 
consequence of the physiological mechanisms involved in fal- 
setto or are due to other factors. It seems most plausible to 
suggest that the physiological mechanisms which relate to 
falsetto also restrict the vocal intensities that can be 
produced . 

Table 5 presents the summary of the analysis of vari- 
ance performed on the vocal intensity data. There are sig- 
nificant F ratios for the main effects of 1.) subject group, 
2) registers, 3) frequencies, and 4) intensity levels. How- 
ever, any interpretation of these main effects is complicated 
by the presence of two significant interaction terms; both 
of which involve registers and intensities. The significant 
three way interaction term involves registers, intensities, 
and frequencies (RFI ) and suggests that at least one two 
factor interaction term differs at one or more levels of the 
third facto r . T h is into r actio n t e r m w as analyz e d by c o m pu t - 
ing simple two way interaction F tests at each level of a 
third factor. The results (shown in Table 17, Appendix C) 
indicate that the register by frequency interaction was 



4s 



Table 5. Summary of the analysis of variance performed on 
vocal intensities measured at each of three frequency levels 
in. the modal and falsetto registers for the naive subjects 
(N - 8) and the singer subjects (N =-, 8). 



Source 



df 



MS 



Betwee n 



G (Subject Groups 
S (Subjects w i t h i n 
Groups) 



1 

14 



1325 .8 



256 . 5 



5.17* 



V.' i t h i n 








R (Registers) 

GR 
RS 


1 
1 

14 


58 71 . 2 
55.5 
38 . 4 


155 . 38* 
1 . 44 


F (Frequencies) 


2 


916.3 


22. 23* 


GF 
FS 


2 

28 


10 . 7 
41 . 2 


. 26 


I (Intensities) 


2 


11753.2 


248 . 8 2* 


GI 
IS 


2 

28 


47.2 
47 . 2 


1 . 00 


RF 

GRF 

RFS 


2 
2 

28 


4 .9 

101.0 

34 . 8 


.14 
2.91 


RI 

GRI 

RIS 


2 
2 

28 


9 26 . 9 
77 .6 

37.4 


2 4.80* 
2 . 07 


FI 

G.FI 

FIS 


4 

4 
56 


6 . 9 
18.1 
23 . 1 


. 3 
. 78 


RFI 


4 


63.0 


6.34 + 


GPFI 
R F I. 3 


4 

56 


8 . 7 

9 . 9 


.97 



*Signi f i can t at the .05 level 



49 



significant only at the minimum intensity condition. Further 
evaluation of this interaction revealed that the intensities 
between the modal and falsetto registers were statistically 
different at all frequency/intensity combinations except at 
the minimum intensity condition for the 25% and 50% frequency 
levels. This result suggests that no cliff eren ces exist between 
the modal and falsetto registers at the minimum intensity 
conditions produced at the 25% and 50% points of the range 
of register overlap; whereas differences occur at the other 
intensity levels and frequencies. 

The significant register times intensity interaction 
term suggests that the intensities measured at the three 
intensity levels were different in the two registers. Simple 
effects tests were conducted (see Table 18, Appendix C) and 
revealed significant F ratios at all intensity levels in the 
two registers. However, inspection of the sums of squares 
from these tables reveals that the differences between the 
modal and falsetto registers were much greater at the most 
comfortable and maximum intensity levels than at the minimum 
intensity level. It would appear that the minimum intensi- 
ties of the two registers are very similar. But as greater 
and greater intensities are produced in the two registers, 
the difference between the two registers, in terms of in- 
tensity, becomes even greater. 



50 



Surprisingly, little information about vocal intensi- 
ties in either the m odal register or the falsetto registe r 
has been reported in the literature. Stout (1944) reports 
vocal intensity ranges for two singers who produced phonations 
within the frequency range of 87 to 440 Hz. Although Stout 
does not identify the register in which the phonations were 
produced, it will be assumed that they were produced in the 
modal register. Furthermore, he does not indicate the refer- 
ence level he used for the dB values he reports. Therefore, 
any c ornpar i s i on s between this study and his must be somewhat 
superficial. In general, the Stout data suggest that the 
magnitude of vocal intensity increases as f increases; a 
similar trend was noted in this study. Furthermore for the 
range of frequencies similar to both studies (196 to 330 Hz), 
the actual range of vocal intensities found in the two studies 
are very similar. However, in the Stout data, the range of 
the vocal intensiti.es became larger as f was increased. In 
this study, on the other hand, the modal register intensity 
ranges of the singers tended to become smaller as f was 
increased. These apparently opposite trends may reflect the 
differences of sample sizes used in the two studies. In this 
study, there were eight singers, selected for homoge-.ity of 
phonational range whereas in Stout's study, only two singers 
w ere used. 



51 

To summarize the results of the intensity procedure, 
it may be indicated that the intensity range of falsetto 
register phonations is much smaller than that for modal 
register. However, at the minimum sustainable intensity con- 
dition, the intensities observed were very similar in magni- 
tude for the two registers. Finally, the results indicate 
that the intensities produced in the two registers tend 
to vary with changes cf the fundamental frequency of phona- 
ti on . 



me 



£e.H£-ej2i.U£i_l stud i e s 

The purposes of these procedures were to determine 
if observers could perceptually distinguish between phona- 
tions produced in the modal and falsetto registers. In or 
procedure, each subject's modal and falsetto register pho- 
nations at the same frequency were paired in every possible 
way, and the observers were asked to respond to the two voice 
qualities as simply same or different (hereafter this pro- 
cedure wil] be referred to as the discrimination procedure). 
In another procedure, each phonation at a specific frequency 
leveJ was presented to naive and experienced observers who 
were instructed to place an equal number of samples into each 
of two undefined categories (hereafter called, the forced 
choice procedure). Finally, all phonations were intermixed 



52 

with no restriction placed on frequency and played to 13 naive 
observers who placed each phonation into one of two undefined 
categories (hereafter called the free choice procedure). 

Register discrimination . Table 6 presents the pro- 
portion of correct same/different responses for each subject 
group and each frequency level by the naive and experienced 
observer groups. The results of the analysis of variance 
performed on the arcsine transformations of the data 5 are 
shown in Table 7. Preliminary tests had revealed no differ- 
ence between the number of correct responses given by the 
three experienced observer groups. Consequently, the data 
from the three experienced subgroups were pooled in the anal- 
ysis. The two main effects (subject groups and frequencies) 
exhibit F ratios that are statistically significant. This 
finding may be interpreted to mean that the number of correct 
discriminations was larger for phonations produced by the 
singer subjects. In addition, the proportion of correct 
responses varied as the fundamental frequency of the phona- 
tion varied, suggesting perhaps that the quality of a register 



A major assumption in an analysis of variance is 
that the mean and the variance of the distribution are inde- 
pendent of each other. When the data are in proportion, the 
mean and the variance are related. The arcsine transforma- 
tion- removes the relationship between the two parameters 
(Winer , 1962) . 



53 



Table 6. Proportion of correct responses as same-different 
for phonations produced in the modal and falsetto registers 
at three frequency levels (208, 26 2,- and 330 Hz) by singers 
and non-singers. Fifteen naive and 15 experienced observers 
rated each sample. 



Subject Groups 



Observer 
G r oups 



Naive 
20 8 26 2 3 30 



Singers 
208 262 330 



Naive 

(N = 15) 



99 .88 



.96 .96 .87 



Speech 

Pathol, ogi s t s 
(N = 5) 



94 .93 .79 



99 1.00 .9. 



Ex per i mental 
Phonet i ci an s 
(N - 5) 



99 .90 .82 



96 .97 . 95 



Singe r s 
(N = 5) 



98 .87 . 83 



99 .97 . 93 



54 



Table 7. Summary of the analysis of variance performed on the 
arcsine transformations of the pr o por t i on of correct responses 
in the paired comparison perceptual procedure. Fifteen naive 
and 15 experienced observers rated the samples. 



Source 



df 



MS 



Bet ween 



G (Observer Groups) 
(Observers Within 
Groups ) 



28 



.02 

.19 



06 



Within 



S (Subject Groups) 

GS 

SO 



1 

1 

28 



2.96 
. 21 
.05 



61 . 86* 
4 .35* 



F (Frequencies) 

FG 

F0 

SF 

SFG 

SFO 



2 

2 

56 

2 

2 
56 



2.84 
.05 
.04 

1 . 35 
.03 
.06 



68.15* 
1 . 36 

21 . 91* 
- 51 



*Significant at the .05 level 



5 5 



changes slightly with a change of fundamental frequency. 
However, precise interpretation of the main effects is not 
possible since two interaction terms are also significant. 
The first term suggests that the naive observers discriminated 
the phonations in a slightly different manner than did the 
experienced observers. Simple effects F tests were conducted 
for this interaction and are shown in Table 19, Appendix C. 
The results suggest that the observer groups tended to dis- 
criminate the phonations of the two subject groups in a dif- 
ferent way. The sums of squares of the experienced observer 
group attained a greater magnitude than the sums of squares 
computed for the naive observers. The data shown in Table 
6 suggest that the naive observers discriminated the pho- 
nations produced by the naive subjects slightly better than 
did the experienced observers. However, the experienced 
observers discriminated phonations produced by the singer 
group slightly better than did the naive observers. Perhaps, 
the two observer groups utilized a slightly different crite- 
rion to determine the similarities or the differences of 
voice quality. If so, the experienced observer group chose 
a criterion which resulted in a higher proportion of correct 
responses to the singer phonations whereas the naive observers 
chose criteria which favored the naive subjects. 

The S x F interaction term shown in Table 7 suggests 



56 



that the phonations produced by the two subject groups were 
judged differently with respect to the frequency level at 
which they were produced. Figure 4 demonstrates that, for 
naive subject phonations, the modal/falsetto distinction 
was differentiated best at 208 Hz whereas the smallest pro- 
portion of correct responses occurred at 330 Hz. 

The test for the simple main effects, shown in Table 
19, Appendix C, indicates that there was no difference between 
the correct number of responses to phonations produced by the 
two subject groups at 208 Hz. At the other frequencies how- 
ever, the two subject groups differ as shown in Figure 4. It 
may be that singers produce different acoustic signals in the 
two registers consistently at all three frequencies whereas 
naive subjects tend to produce the clearest acoustic difference 
at the low frequency of the range of register overlap. 

The results of the analysis of variance suggest that 
pooling the subject groups and/ or frequencies would be un- 
tenable. Thus tests of the significance of the proportions 
shewn in Table 6 were conducted at each frequency level for 
the two subject groups. The results of these tests revealed 
that both observer groups discriminated the phonations produced 
by the two subject groups better than chance (i.e., p = .50). 
Therefore, it appears that modal and falsetto register phona- 
tions can be distinguished perceptually. It is interesting 



57 




203 



232 
FREQUENCY 



( 



Hz} 



Figure 4. Proportion of correct responses 

in the paired comparison perceptual procedure 

to pronations produced at three frequency levels 

by the naive (x-x) and singer (0-0) subjects. 



58 



to note that the naive observers tended to discriminate 
the naive subject phonations slightly better than they 
discriminated the singer phonations. On the other hand, 
the experienced observers tended to discriminate the singers* 
phonations slightly better than they discriminated the naive 
subjects' phonations. 



Register categorization 

Forced choice procedure . In this procedure, naive 
and experienced observers were required to classify each 
sample with respect to one of two categories labelled A or B. 
Each observer was allowed to establish his own judgmental 
criteria, although he was instructed to judge each sample 
simply on the basis of voice quality. As stated previously, 
the response sheet of each observer was inspected to ascertain 
the particular label (i.e., A or B) used to categorize each 

phonation. The proportion of correct responses was then com- 

6 

puted . 

Since each observer heard each phonation twice, it is 
possible to estimate the reliability of the group's responses. 



6 
The proportion of correct responses shown in Table 11 

represents the proportion of correct judgments for only one o: 

the two registers. Since the observers were forced to render 

10 A judgments and 10 B judgments, an error in one category 

also results in an error in the other category. Therefc 

both registers will have the 

men t s . 



ore 



te same proportion of correct judg- 



59 



Table 8 presents a joint distribution of modal and falsetto 
register responses for the two presentations of the stimuli 
by the naive observers. Table 9 presents the identical sum- 
mary for the three subgroups of experienced observers. E a ch 
cell represents the responses to the phonation when he first 
heard the sample and the judgment when he heard it the second 
time. For example, the modal/modal cell means that when an 
observer heard the phonation each time, he labelled it modal 
on both occasions. The modal/falsetto cells indicate that when 
the observer heard, the sample the first time he labelled it 
modal, but the second time he heard it he labelled it falsetto. 
A high number of identical responses (i.e., modal/ modal and 
falsetto/falsetto) would suggest that the observers were 
reasonably consistent in their assignment of labels to a pho- 
nation. I n order to evaluate the proportion of similar 
responses, phi coefficients were computed from the data 
shown in Table 8 and Table 9. The naive group phi coef- 
ficient was .31 which is significant at the .05 level of 

8 
confidence. The speech pathologists, experimental phoneti- 
cians and singers obtained phi coefficients of .30, .25, and .31 
respectively; all of which are significant at the .05 level. 



The test statistic used to evaluate phi is N0 2 which 
is distributed as X 2 with one degree of freedom. 



60 



Table 8. Distribution of modal and falsetto responses on 
the first and second, presentation of samples produced by 
naive and singer subjects in the forced choice perceptual 
procedure. Fifteen naive observers rated each sample 



First Presentation 
Modal Falsetto Total 





a 









•n 




+> 


■e 


Cti 


a 


+> 





C 


o 


0! 


0) 


(0 


CO 


d) 




Sm 




ftl 



Modal 

Falsetto 

Total 



196 



109 



295 



98 



207 



305 



294 



306 



600 



61 



Table 9 



Distribution 



n of modal and falsetto responses on 
the first and second presentation of samples produced by 
naive and singer subjects in the forced choice perceptual 
procedure. Experienced observers consisting of sppech 
pathologists CN = 4), experimental phoneticians (N = 5) 
and singers (N = 5) rated each sample. 



First Presentation 
Modal Falsetto 



Total 



S_p.e e ch Pat ho lo g i s t s 



a 
o 

•H 
+> 

Si 

£ 
0) 
en 

U 
Ck 

TS 

c 

o 
o 

0) 

ra 



Modal 

F a Isetto 



Experimental 
Pjjon et ici an s 

Mo d a 1 
Falsetto 



Singer s 

Modal 
Fal set to 



78 

42 

120 



94 

56 

150 



98 

52 

150 



42 

78 

120 



56 

94 

150 



52 

98 
150 



120 
1 20 
240 



150 
1 50 
300 



150 
1 50 
300 



62 



Although not conclusive, these results suggest that the 
observers rated the phonations in a similar fashion each 
time they were presented. 

Further evidence that the observers rated the samples 
in a similar fashion on each presentation is provided by the 
results of the t-tests performed on the difference between 
the number of correct responses on the first presentation and 
the number of correct responses on the second presentation. 
All t-tests were non-significant which suggests that the 
observers rated the phonations in a similar fashion. Conse- 
quently, the'number of correct responses from both presenta- 
tions was pooled and subsequent statistical analysis was per- 
formed on these pooled data. 

Table 10 presents the summary of the analysis of variance 
performed on the arcsine transformations of the proportion data. 
Preliminary tests of the data had indicated that the three 
experienced observer groups could be pooled in this analysis. 
The only significant F ratio shown in this analysis is the 
three way interaction term. Apparently, the two observer 
groups categorized the phonations from the two subject groups 
and the three frequency levels in different ways. The simple 
interaction F tests conducted at each level of the third 
factor (Table 20, Appendix C) resulted in a significant two 
way interaction between observer groups and frequency level 



63 



Table 10. Summary of the analysis of variance performed on 
the arcsine transformations of the proportion of correct 
responses for three frequency levels and two subject groups 
in the forced choice perceptual procedure. Fifteen 
and 15 experienced observers rated the samples. 



naive 



Source 



df 



MS 



Between 



G (Observer Groups) 
O (Observers Within 
Groups) 



I 
28 



00 



09 



04 



Within 



S (Subject Groups 

GS 

SO 



. 52 

00 

.14 



60 
03 



F (Frequencies) 

GF 

F0 



2 

2 
56 



21 



07 



2. 90 



62 
62 



SF 

GSF 

SFO 



2 
2 

56 



02 

64 
18 



12 
60* 



♦Significant at the .05 level 



64 



only for the naive subject group. This result is also noted 
in Figure 5 which plots the data by the observer groups for 
each subject group at each frequency. Large reversals of the 
proportion of correct responses are noted for the naive sub- 
jects phonating at the three frequencies whereas the singer 
phonations were categorized by the two observer groups in a 
more consistent manner. 

Table 11 presents the proportion of correct responses 
at each frequency level for each subject group. The signifi- 
cance of the proportions shown in Table 11 was evaluated by 
computing a critical proportion required to obtain a t value 
equal to a t at the .05 level. The proportions which exceeded 
this critical value are starred in Table 11. All proportions, 
with one exception, of the two pooled groups (naive and ex- 
perienced) reached significance. For the experienced observer 
subgroups, 11 of the 18 proportions are statistically signifi- 
cant. Interestingly, most of the singer phonations were 
categorized correctly more often than the naive subject pho- 
nations by these subgroups of observers. Perhaps, singers 
produce a more obvious voice quality difference between the 
two registers, thus enabling naive and experienced observers 
to correctly categorize a phonation. 

The results from the individual experienced observer 
subgroups reflects the variability of individual register 



6 5 




I 
O 

q 

C 1" c w o 

S ~: o IM %.' 



O 

o 



r 

o 


i 
o 


1 

o 


o 


N 


w 


03 


\J V! Q 


JO K 



o 
!0L1.H0( 



JftHrl 



CO 








B 


1 


K 




Ti 




•H 


O 


liJ 




b 


<D 


X! 


u 


CD 




a 


.B 


P 


ft 








P 


•H 





















/■"■\ 




^ 


r — i 


CO 




X 


>> 




cti 






1 


JQ 


to 


3 






X 




H 


P 






^ 


■3 





ft 








0) 


> 





> 

< 




(I) 


O 










> 

•H 
Ctj 


3 
ti 




H 




ft 


«s 




fi 


ft 


O 

B 









0) 














J! 


cn 


3 


■H 






•p 


d 


cr 


O 














r t* 


CO 


co 


>> 


•H 


^ 


O 


cr 


D. 


fit 


■P 


<H 




w 


15 




cd 




■3 


o 


O 


cn 


C 








2: 
















tn 


,B 


J-l 


h 


CO 


O 


B 


a 


-B 


O 











P 


<H 




}- 


ft 


■a 








O 


tn 


CD 


<H 







bj 





P 







U! 


-3 


U 


cd 




P 


> 

< 


Q 
CO 


-P 
O 





O 

(0 


B 
•H 


2 





xs 











h 


=c 




a 






u 




P 


cd 









cn 


n3 


H 









r M 




H 


CO 

K 




<h 


CD 

> 


cn 

p 



> 







f-t 
CD 













B 




43 




5-i 




CO 




•H 





3 


4-> 






P 




cn 


cn 






Sh 


/-\ 




•H 












^ 


bfl 






a 


1 


















tsO 


Sh 


LsJ 




h 


^-* 


r] 




^" 




ft 




•H 


<w 


< 






CD 


cn 


O 


*■• 




. 


O 


■c 









co 


fi 



c 
cti 


B 






0) 


■H 




CO 






?H 


h 





h m 






3 





> 


3 






bo 


ft 


■H 


" 






•H 


X 


cd 


X! 






Ft, 


CD 


B 


■P 



66 



Table 11. Proportion of correct categorizations of the modal 
and falsetto register phonations produced by singers and naive 
subjects at each of three frequency Levels within the range of 
register overlap in the forced choice categorization procedure 
Fifteen naive and 15 experienced observers judged each phona- 
tion. The experienced observers were subdivided into three 
groups as is shown in the table. 



Observer 
Groups 



Naive 
208 262 330 



Singers 
208 262 330 



Naive 



61* 



63 ; 



60 : 



70* . 63* . 62* 



S peech 

Pathol ogi s t s 



7 5* 



60 



64* .64* .53 



Experimental 

Phoneticians 



72* .54 .60 



6 8 * .68* .76* 



■mger s 



70* .54 .58 



68* .68* .80* 



Ex peri enc ed 
b s e i' v e r s 
(Pooled Across 
Subgroups ) 



7 2* .52 .59* 



.67* .67* .70* 



*Significant at the .05 level. 



67 



judgments and partially demonstrates the need for a large 
number of observers to obtain stable judgments of voice 
quality. With a small number of judges, one individual 
can adversely (or favorably) affect the overall group 
proportion. Occasionally, such individual effects were 
observed in this study. Perhaps with a large N in each ex- 
perienced observer subgroup, the proportion of correct re- 
sponses would have been higher. In spite of the small 
number of observers in each subgroup, the phonation conditions 
were still correctly categorized better than at the chance 
level . 

One might conclude, therefore, that registers in gen- 
eral can be correctly identified when the observers use a 
voice quality criterion. In view of the lower proportion 
of correct responses, it would appear that the observers ex- 
perienced greater difficulty categorizing the acoustic 
signals of the phonations produced in the two registers than 
they did discriminating between the two signals. This in- 
creased difficulty may merely reflect the more complicated 
perceptual task the observers were required to perform. 

In summary, the data suggest that, at a simple level 
of perceptual processing (discrimination), observers can 
distinguish the voice quality of the modal and falsetto 
registers extremely well. The observers were also able to 



68 



identify and categorize the two register voice qualities 
although the proportion of correct responses is not as high 
as for the discrimination task. Nevertheless, it is apparent 
that the voice quality associated with the modal register 
can be reliably classified and differentiated from the voice 
quality of falsetto register phonations. 

Free choice procedure . In this procedure the observers 
were not limited to responding v/ith an equal number of A and 
B labels to the phonations. In addition, the three frequency 
levels were intermixed throughout the tape. The observers' 
responses for the two presentations of the stimuli were 
analyzed in order to estimate the within observer reliability. 
Table 12 presents the results of this analysis. As an index 
of observer agreement, a phi coefficient of .46 was computed 
for these data which is significant at the .05 level. Although 
this result suggests that the 13 observers rated the stimuli 
in a similar fashion, the agreement of the observers with 
themselves is not as high as would be desirable. 

The results of the t-tests performed on the difference 
between the number of correct responses of the two presenta- 
tions revealed a significant t at t wo " phonat i on conditions, 
vi_z. , the naive 330 Hz modal register condition and the singer 
262 falsetto register condition. In view of these two dif- 
ferences in the observer judgments, each presentation of the 



6 9 



Table 12. Distribution of modal and falsetto responses on 
the first and second presentations of samples produced by 
naive and singer subjects in the free choice perceptual 
procedure. Thirteen naive observers rated each sample. 



First Presentation 
Modal Falsetto Total 



a 

















■H 

-p 


Modal 


302 


1 25 


■O is! 








C -P 








o c 

o 1> 


Falsetto 


100 


240 


(D E 








W CD 


Total 


402 


365 


P-i 









427 
340 
767* 



*0ne ohonation condition was eliminated thus result 
ing in 767 joint responses. 



70 



stimuli was treated separately in the subsequent analysis of 
variance. In general, it appears that the observers were 
much more variable in the categorization of the samples 
presented in this procedure. The data, therefore, should be 
interpreted with caution. 

The results of the analysis of variance performed on 
the first responses are shown in Table 13 whereas Table 14 
presents the results of the analysis on the second presenta- 
tion data. Again, the results of these analyses illustrates 
the difference between the first and second presentation 
responses. In Table 13, a three-factor interaction term is 
significantly different from chance whereas in T a ble 14, no 
interaction term reached significance. Figure 6 illustrates 
how the proportion of correct responses varied as a function 
of frequency level for the first presentation responses. It 
is apparent that the voice quality judgments of the observers 
in this procedure were influenced by the fundamental fre- 
quency of the phonation and by the subject who originally 
produced the phonation. Evidence for this latter statement 
is provided by the results of the simple effects tests per- 
formed on the two interaction terms (see Table 21, Appen- 
dix C). The simple two-way interaction between the subject 
groups and the two registers was significant at 262 Hz and 
330 Hz but non-significant at 20S Hz. This result means 



71 



Table 13. Summary of the analysis of variance performed on 
the first responses to modal and falsetto register phonations 
produced at three frequencies by naive and singer subjects 
in the free choice perceptual procedure. Thirteen naive 
observers rated the samples. 



S our ce 



Between 

(Observers) 



df 



12 



MS 



2. 77 



Within 

S (Subject Groups) 
SO 



R (R 


egi s t er s 


RO 




F (F 


r equenci es 


FO 




SR 




SRO 




SF 




SFO 




RF 




RFO 




SRF 




SRFO 





1 

12 

1 
12 

2 

24 

1 
12 

2 
24 

2 
24 

2 
24 





. 10 


1 


. 53 


1 


. 64 


1 


. 27 


5 


. 62 


2 


. 54 




10 


1 


01 




04 


1 


18 


20 


51 


3 


05 


6 . 


54 




95 



07 



1 . 30 



2. 21 



.03 



.04 



6 . 73* 



6 .91* 



*Significant at the .05 level 



72 



Table 14. Summary of the analysis of variance performed on 
the second responses to modal and falsetto phonations pro- 
duced at three frequencies by naive and singer subjects in 
the free choice perceptual procedure. Thirteen naive ob- 
s e r vers rated t h e s am pi e s . 



Source 



df 



MS 



Between 



(Observers) 



12 



2.13 



Within 



S (Subject Groups) 
SO 

R (Registers 

RO 

F (Frequencies) 
FO 

SR 
SRO 

ST 
SFO 

RF 
RFO 

SRF 
S R F O 



1 
12 

1 
12 

2 
24 

1 
12 

2 

24 

2 

24 

2 

24 





. 3.1 


1 


- 37 


25 


. 44 


1 


.89 


8 


.95 


2 


. 5 3 




16 


1 


16 




79 


1 


41 


9 


00 


2. 


92 




7 9 


1 . 


57 



23 



13 . 48* 



3 . 54* 



. 14 



56 



3 . 08 



- 51 



^Significant at: the .05 level. 



73 




i 
O 

o 



o 



1 

o 

o 


1 
o 


1 

o 
to 


■ I 

o 

to 


1 

o 


1 

o 
to 









CO 


CCJ 










P 


p 













cd 




p 






CD 
•i~5 


Q 




J— 







,Q 






1- 




p 


3 






LU 




p 


CO 


a 




CO 




(1) 




cd 




_J 




CD 


/-N 


H 




< 




H 


o 


h 




u. 




esj 


1 


CD 








<H 


o 


> 










•a 








_J 




S3 


u 


U 




< 




a) 


<u 


CD 




a 






b£ 


P 




o 




H 
Ctj 


■H 


CO 
•H 




« ' — 







CO 


to 

CD 






CO 


s 


•a 


^ 




o 
(- 


o 



p 


cd 







J— 

LU 


h- 


ID 




ai 


■H 


CO 


D 


a> 


i 


M 


r-i 


-J 


"■~J> 


10 


X 


£ 





<^ 


o 


a 


^ 


Si 


g 


u. 


o 





0) 


*4 


P 






tn 


> 


o 


CO 




rx 


CD 


•H 


£?> 




.J 


Ui 


U 


cd 


P 


CD 


< 
O 


CO 
Ci> 


+> 

o 
CD 


c. 





O 


•m 




U 




CO 


P 




LU 


u 


-a 


iH 






cc 





0) 


CD 


CD 


O 







V 


►> 


o 


;•-• 
H 




<H 


s 

T3 



i-i 


CO 
C! 




o 










L'J 






Sh 


>i 


a 


CO 




a 


Q 


o 


CO 


_J 









rl 


CD 


< 




■H 


CO 


CD 


h 


U. 




P 


r^ 


2 








fc 


c 


C"' 


^J 









•H 


CD 


CO 






a 


P 


%, 


M 


_J 







CO 


'i-i 


■- 1 


< 




^ 


^ 




<H 


Q 




£h 





CD 




O 






r*> 


CD 


CD 














c - 








.^J 


P 






CD 


CD 


+ j 


g 






CD 


p 


CD 









^ 


CO 


,£ 


SU 






2 


■H 


P 


<H 






fan 


t5 




CO 






•H 


CD 


-p 


^ 



O V •'' •> : A J C "3 v . ' ! A "i •■ ; \ " A A 1 A !> ! A i ! U A ..' A 5J ,4 

vi Zi o i . v/ c a . , a JLJ .:< cJ cj U J .:■ O hUUc U c U a cl 



74 



that the interaction between subjects and registers was 
present at the higher frequency levels but not at the low 
frequency level. This is also shown in Figure 6. Further- 
more, the register by frequency interaction term (RF) was 
not significant at 262 II z but was significant at 208 Hz and 
330 Hz. Apparently, the observers utilized fundamental fre- 
quency as a partial criterion by which to judge these samples. 
This is shown in Figure 7 which is a plot of the second re- 
sponse data. Since falsetto register phonations are normally 
produced at a relatively high fundamental frequency, this 
criterion resulted in a higher preparation of falsetto re- 
sponses when the fundamental frequency (or perceptually, pitch) 
was high. When the fundamental frequency (pitch) was low, the 
observers tended to judge the pronation as representative of 
the modal register. 

Table 15 presents the proportion of correct responses 
in the free choice procedure for each presentation of the 
stimuli. Six proportions attained a value required for 
significance at the .05 level. It is noted that all propor- 
tions at the 2 S Hz modal register condition and three of the 
four proportions at the 330 Hz falsetto condition are statis- 
tically significant. Only two of the remaining 16 proportions 
were significant. This result would suggest that the funda- 
mental frequency of the phonation influences the observers 



75 




O 



1 


i 


\\ 


o 


o 





4-> 

w 

•H 

hi) 

U 

U xi 

■P 
O 

+■' -P 

-p cd 
© 

W en 

H -P 

cd u 

<H 

■n 

TS ft 

5 3 
cd to 

H /->. 

cd O 

xs I 

o o 

6 — 



o 

4-> 

cn 

(1) 

cn 

a 
o 
ft 



c£ 
•H 

■d 



0) 

o 

■H 
O 

tt> 

.a 

CD 4-> 


Sh <H 

«M O 

a 

* o 

■P -H 

C cd 

ft 

rd cn 

r-t 

U U 

& 
> 

o -o 

u o 



-P 

tfl cn 

Si 

JZ 

h -P 



co cd 




o 



u 

o 
o 

o 





I 


> 

■H 

cd 
c 

■a 

o 
o 
3 
■rJ 

o 



SE 



• ?■! <A -' O V ; V I A 13 V M A •"•, ,t A f. | A I I \ ! A .-' A VJ 



•rf 

4-> 

m cd 

zi a 

CD 

■r-l Si 

h ft 



o o 

'-P 
fcfl 

C 

cd h 

h * 

cd 

xi -p 

-p cd 
Q 
G 

•H 

+J 

■H 5-i 

■d 

n 

-i o 

o 

> f* 



>, cd 

u 3 * 

,11 ft -J 



+J 
O. Cfl 



76 



Table 15. Proportion of correct responses for phonations 
produced in the modal and falsetto registers by naive and sin- 
ger subjects at three frequency levels In the free choice 
perceptual procedure. Thirteen naive observers rated each 
sample twice. The results are reported separately for each 
presentation (see text). 



Subject 
Groups 



Naive 



Naive 
209 262 330 



Singers 
208 262 330 



First Response 
Second Response 



6 3* .6 5 ; 



6 2* .60 



46 



58 



40 



31 



43 



82* 



3 7 .68* 



Sineers 



First Response 
Second Response 



68* .48 .58 



65* .52 



6 2"' 



35 .55 



3 5 .34 



68 



57 



♦Significant at the .05 level 



77 



judgment of register voice quality. Consequently, it 
appears that the fundamental frequency and voice quality 
of a ph.onati.on produced in a particular register interact 
and serve as dual cues which the observers utilize in clas- 
sifying these ph on at ions. 



Im plications of the pe rce ptua l studies 

The significance of the three perceptual procedures 
conducted in this study is that they demonstrate the levels 
at which observers can correctly distinguish the modal and 
falsetto registers using voice quality as a criterion. First, 
with no immediate training in the rating of these phonations, 
the observers can differentiate the acoustic signals of the 
two registers. This result suggests that phonations produced 
in the modal and falsetto registers are different acoustically. 
At the next level of perceptual processing, the observers 
classified the same phonations. When provided with a restricted 
set of samples, they are able to place the phonations into the 
proper categories at a level significantly different from 
chance. However, when asked to rate phonations which vary in 
f Q in addition to voice quality, they tended to categorize on 
the basis of f Q and voice quality. Therefore, it would appear 
that the modal and falsetto registers are not identified solely 
by a specific voice quality. This statement does not mean 



78 



that these registers lack a characteristic voice quality; 
indeed, the data support the hypothesis that there is unique 
voice quality. However, it is also recognized that the voice 
quality of phonations produced in the same register varies 
from subject to subject and also varies at different funda- 
mental frequencies. Some of these voice quality differences 
may be accounted for by the differences of laryngeal ' si ze , 
vocal fold mass, and vocal fold tension which exist across 
individual subjects. 

Within an individual, registers are probably the 
result of a unique and identifiable physiological state of 
the larynx. Although much is known about the physiology of 
the larynx and some data are available about laryngeal 
physiology in different registers, it is not possible at 
this time to identify that physiological state (in terms of 
vocal fold length, glottal area, vocal fold thickness and 
air stream dynamics) which can be associated either with the 
modal or the falsetto register. But if such a state exists, 
it is obvious that acoustically it will result in some combi- 
nation of f , vocal intensity and voice quality. One such 
combination might enable the listener- to identify one register 
whereas another combination may permit him to identify another 
register. Normally, it is this combination that the 
observer utilizes to distinguish registers. However, 



79 



when f Q and vocal intensity are controlled, the observer 
can utilize only one parameter, i.e., voice quality, to 
distinguish the two registers. It would seem, therefore, 
that voice quality is a very important feature of a reg- 
ister pronation. However, both f and vocal intensity seem 
to serve an ancillary role in the production and perception 
of vocal registers. 

Spectrum measurements 

The results of the harmonic analysis performed on 
each phonation are displayed in Figure 8. On the ordinate 
is plotted the number of harmonic partials which possessed 
measurable energy in each sample. 8 The mean number of 
partials for each subject group is shown separately. Also 
shown are the number of harmonics at each frequency when 
the data f m the two subject groups were pooled. One 
trend suggested in this figure is that modal register pho- 
nations possess energy in a greater number of harmonics 
than do falsetto register phonations. In other words, the 
frequency extent of a modal register phonation is wider than 
the corresponding frequency extent of falsetto register pho- 
nations. This observation is in agreement with the previous 



8 

A harmonic was considered measurable if it possessed 
energy at least two dB above the noise level. 



8 




10 






o 

Cvl 








S 




IN 


OW 


h 



lO 



N 



o 

«£• 
LU 
Z> 
G 
W 
& 



o 



:o ysawnN 



o 
I 

o 



o 

4-> 

+J 

0! 

m 

«H 

■a 
fl 
cd 



01 

<D 

o 

c 

0> 

cr 

CD 


o 

-P 

+> 
cd 






o 



cd « 

-a 

M 

6 0) 



+-> 

en cd 
0) 

M 0) 

ft > 

i-l 

tn cd 

o a 






ft 

cd 



a 

o 

£ U 

h T) 8) 

cd cu +j 

<h -a 
o o 



H 
(1) 



CD 



ft 

en 

a 
o 

•H 

cd 

ft 



o 



to 

CD 



in M -H 

w bo + j 

•H 0) -H 



81 



work of several authors. 

Table 16 presents the results of the analysis of 
variance performed on these data. The trend observed in 
Figure 8 is confirmed by the results of this analysi, 
i.e., the number of harmonic partials differs significantly 
in the phonations from the two registers. 

The average number of partials observed in these 
phonations agrees with the number of partials measured in 
phonations by Sonninen (1962). In this study, the average 
number of partials at 208 Hz was about 20 whereas Sonninen 
reports 30 partials in chest register phonations of his one 
female singer. On the other hand, Baisler (1950) reports 
about five paritals in the modal register phonations of his 
two subjects, and about four partials in falsetto phonations. 
Of course, differences in methodology could account for at 
least some of these differences in results. Moreover, Baisler 
obtained his acoustic records while the subject was phonating 
with a laryngeal mirror in his mouth and high speed motion 
pictures were taken. In the present study, emphasis was 
placed on obtaining high quality acoustic recordings with 
the subjects phonating in as near normal conditions as pos- 
sible. Perhaps these differences reflect themselves in the 



Baisler (1950); Luchsinger and A r nold (1965); Mills 
(1913) ; Smith (19 57) . 



82 



Table 16. Summary of the analysis of variance performed on 
the number of partials measured in the pronations produced 
by naive and singer subjects at each frequency level. 



Sour ce 



df 



MS 



Between 



G (Subject Groups) 
S (Subjects within 
Groups ) 



11 . 3 

177 . 4 



Within 



R (Regi ster s ) 

GR 

RS 



1 
1 
8 



564 . 3 
11 . 3 
57 . 4 



9 . 8* 
. 2 



F (Frequencies) 

GF 

FS 



2 

2 

16 



38 . 5 

3 .6 

12.6 



3 . 1 
. 3 



RF 

GRF 

RFS 



2 

2 

16 



^Significant at the .05 level 



32.6 
6 .3 
9 .4 



3. 5 
. 3 



83 



total number of harmonics present in the phonation. 

Since the number of partials in falsetto register 
phonations is different from the number of partials in modal 
register phonations, one would expect differences in the 
quality of the tones produced in the two registers. Con- 
sequently, the results of the harmonic analysis supports 
the results of the perceptual studies. Since there are 
physical differences in the acoustic signal, the observers 
were able to classify the phonations into two voice quality 
groups. In view of the obvious differences in the number 
of partials 'in the phonations, it is probable that the 
voice qualities of the two registers are primarily related 
to the harmonic spectrum. 



CHAPTER IV 
SUMMARY AND CONCLUSIONS 

The purposes of this investigation were to determine 
1) the range of frequencies that could be produced in the 
modal and the falsetto registers, 2) the range of vocal 
intensities that could be produced in the two registers, 
3) the voice quality associated with each of the two registers, 
and 4) the distribution of the harmonic components in pho- 
nations produced in the two registers. Singers and non- 
singers were obtained from a university population and pro- 
duced the lowest and the highest possible frequency in each 
register. From this larger screening group, eight singers 
and eight non-singers were selected to produce their vocal 
intensity range at each of three frequency levels of the 
range of frequency overlap between the two registers. Finally, 
five subjects from each group produced a sample of the vowel 
/Q./ in the two registers at 208, 262, and 330 Hz. Since every 
subject reproduced his phonational range each time he was 
seen, an estimate of the stability of each phonational range 
could be obtained. 



84 



85 



At each frequency level, 208, 262, and 330 Hz, the 
phonatlons from the two registers produced by one subject 
at one frequency level were paired in every possible way. 
Naive and experienced observers responded to these phona- 
tlons by indicating same or different. In another procedure, 
each phonation from the 10 subjects appeared individually 
and all phonations within a tape were of identical fre- 
quency. Experienced and naive observers sorted these 20 
phonations into two groups of 10 samples each; each group 
representative of a particular voice quality as determined 
by the observer. In a third perceptual procedure, 13 naive 
observers heard each phonation individually and frequency 
level and register were allowed to vary. In both catego- 
rization procedures, each phonation was presented so that 
the consistency of the observer's responses could be eval- 
uated . 

The final procedure entailed an acoustic analysis 
of each phonation presented in the perceptual studies. 
Each sample was played to a wave analyzer in order to ob- 
tain an amplitude by frequency display and thereby gener- 
ate the harmonic spectrum of" the phonation. 

The results of this investigation can be summarized 
as follows. 



86 



Subjects tend to produce a slightly larger 
range of frequencies in the modal register 
than they produce in the falsetto register. 
As a whole, singers produce a slightly larger 
range of frequencies in the two registers 
than do non-singers although the differences 
between the two groups were not statistically 
significant . 

The naive subjects produced a three semitone 
overlap of frequency between themodal and 
falsetto registers whereas the singer pro- 
duced a nine semitone overlap. 

For both subject groups, the phonational 
ranges measured on several separate oc- 
casions were very similar. 

The range of vocal intensities in the modal 
register is larger than in the falsetto 
register; moreover, subjects produce a higher 
maximum intensity level in the m odal . r egi st er 
than they can in the falsetto register. Fur- 
thermore, vocal intensities at one frequency 
within the range of frequency overlap between 
the two registers are not comparable with the 
intensity produced at another frequency within 
the overlap. Singers, generally, can produce 
a greater range of intensity in the two reg- 
isters than can non-singers. 



Phonations produced in the two registers can 
be distinguished on the basis of voice quality. 
With few exceptions, modal and falsetto register 
phonations can be correctly placed into the 
proper categories by observers, solely on the 
basis of voice quality, providing that the 
phonations heard by the observers are produced 
at an identical frequency. When phonations of 
different frequency as well as register were 
heard by the observers, they -tended to cate- 
gorize the phonations on the basis of both 
frequency and to voice quality. 

The number of harmonic partials measured 
in the modal register phonations was greater 
than the number of harmonics measured in 
falsetto register phonations. 



From these data, it is concluded that the modal and 
falsetto registers exhibit frequency, intensity, and voice 
quality differences. Generally, the modal register is 
produced at low fundamental frequencies (80 and 300 Hz) 
with a distinct voice quality and a large number of frequency 
components. The falsetto register is located above the modal 
register on the frequency continuum and exhibits fewer fre- 
quency components. Pronations from the modal register are 
produced with a greater vocal intensity than can phonations 
produced in the falsetto register. Finally, although a 
distinct voice quality is associated with each register, 
the fundamental frequencies also associated with it assist 
in the identification of that register. 



APPENDIX A 



INSTRUCTIONS TO OBSERVERS 
PERCEPTUAL PROCEDURES 



8 9 



INSTRUCTIONS TO THE OBSERVERS 
FOR THE PAIRED COMPARISON PERCEPTUAL PROCEDURE 



You are about to hear 120 pairs of phonations. Lis- 
ten to each member of the pair carefully and then decide if 
the two phonations in the pair sounded the same in voice 
quality or if they sounded different. If they sounded the 
same, darken the S on the response sheet. If they sounded 
different, darken the D on the sheet. Each pair will be 
heard only once. There are a total of 20 pairs for each 
tape and a total of six tapes. There will be a short pause 
between each tape but otherwise there will be no stopping. 
Are there any questions? 



90 



INSTRUCTIONS TO OBSERVERS 
FORCED-CHOICE PERCEPTUAL PROCEDURE 



Ygu are going to hear a tape recording of 20 voices. 
Your task is to divide these samples into two groups; there 
must be 10 samples in each group. Please divide these sam- 
ples on the basis of voice quality, not pitch or loudness, 
but quality. Listen to the tape once and decide which 
voices; will go into group A and which will go into group 
B. When the tape is repeated, place an A each time you hear 
a voice quality you believe should go into group A and a B 
for those voices representative of group B. It does not 
matter which letter you assign to which group so long as 
you are consistent in your judgments. Thus, all qualities 
in group A should sound either the same or similar to each 
other and all qualities in group B should sound similar to 
each other. The tape will be played a third time so that 
you may check your responses. At the conclusion of the first 
tape of 20 voices, another tape will be played with 20 more 
voices. Please follow the same directions. Finally, a 
tlii r d t a | ) e w i 11 be p r e s ente d w i t h 20 m o r e voic e s . Each tape 
v.- ill be clear! y identified. 



91 



INSTRUCTIONS TO OBSERVERS 
IN THE FREE CHOICE PERCEPTUAL PROCEDURE 



You will shortly hear a tape recording of 120 voices, 
all of the vowel / Q / as in father. Your task is to place 
each sample into one of two categories, the characteristics 
of which you will determine. The criterion that you will use 
to establish these categories is voice quality. Please do 
not allow pitch or loudness to interfere with your judgments 
of voice quality. 

Before playing the 120 samples, a short practice ses- 
sion will be held, consisting of 30 samples. Actually, these 
will be 10 samples repeated three times. The first time you 
hear the samples, listen carefully and try to determine in 
which one of two general voice quality categories a phonation 
can be placed. The second time you hear the samples, place an 
A for those samples you think are representative of the voice 
quality A or a B for those samples which are representative 
of the category B. The third time you hear the samples, 
please check your responses and make any changes that are 
appropriate. Are there any questions? 

(Play practice samples.) 

Are there any questions prior to the presentation of 
the experimental samples? 



APPENDIX B 



RESPONSE SHEETS 
PERC E PT UAL PROC ED URES 



Name 



RESPONSE SHEET FOR THE PAIRED 
COMPARISON PERCEPTUAL PROCEDURE 



Condition 1 



Condition 3 



1 . 


S 


D 


2. 


S 


D 


3 . 


s 


D 


4. 


s 


D 


5. 


s 


D 


6 . 


s 


D 


7. 


s 


D 


8 . 


s 


D 


9 . 


s 


D 


10 . 


s 


D 


11 . 


s 


I) 


1 2 . 


s 


D 


13 . 


s 


D 


14 . 


s 


D 


15. 


s 


D 


16 . 


s 


D 


17 . 


s 


D 


18 . 


s 


D 


19 . 


s 


D 


20 . 


s 


D 



1 . 


S 


I) 


2. 


S 


D 


3 . 


S 


D 


4. 


S 


D 


5 . 


s 


D 


6 . 


s 


D 


7 . 


s 


D 


8 . 


s 


D 


9 . 


s 


D 


10 . 


s 


D 


11 . 


s 


D 


12. 


s 


D 


13 . 


s 


D 


14 . 


s 


D 


15 . 


s 


D 


16 . 


s 


I) 


17 . 


s 


D 


18 . 


s 


D 


19 . 


s 


D 


20 . 


s 


D 



Condi 


t i on 


5 


1 . 


S 


D 


2 . 


S 


D 


3 . 


s 


D 


4. 


s 


D 


5 . 


s 


D 


6 . 


s 


D 


7 . 


s 


D 


8 . 


s 


D 


9 . 


s 


D 


10 . 


s 


D 


11 . 


s 


D 


12. 


s 


D 


13 . 


s 


D 


14 . 


s 


D 


15 . 


s 


D 


16 . 


s 


D 


17 . 


s 


D 


18 . 


s 


D 


19 . 


s 


D 


20 . 


s 


I) 



Condition 2 



Condi ti on 4 



Condi ti on 6 



1 . 


S 


D 


2. 


s 


D 


3. 


s 


D 


4. 


s 


D 


5. 


s 


D 


6 . 


s 


D 


7 . 


s 


D 


8 . 


s 


D 


9 . 


s 


D 


10 . 


s 


D 


1 ] . 


s 


D 


12 . 


s 


D 


13 . 


s 


D 


1 4 . 


s 


D 


1 5 . 


s 


D 


16 . 


s 


D 


17 . 


s 


D 


18 . 


s 


D 


19 . 


s 


D 


20 . 


s 


D 



1 . 


S 


D 


2. 


S 


I) 


3 . 


s 


D 


4. 


s 


D 


5. 


s 


D 


6 . 


s 


D 


7 . 


s 


D 


8 . 


s 


D 


9 . 


s 


D 


10 . 


s 


D 


1] . 


s 


D 


12 . 


s 


D 


13 • 


s 


D 


1 4 . 


s 


D 


15 . 


s 


D 


16 . 


s 


D 


17 . 


s 


D 


I 8 . 


s 


D 


19 . 


s 


D 


20. 


s . 


D 



1 . 


S 


D 


2 . 


s 


D 


3 . 


s 


D 


4. 


s 


D 


5- 


s 


D 


6 . 


s 


D 


7 . 


s 


D 


8 . ■ 


s 


D 


9 . 


s 


D 


10 . 


s 


D 


11 . 


s 


D 


12. 


s 


D 


1 3 . 


s 


D 


14 . 


s 


D 


I 5 . 


s 


D 


16 . 


s 


D 


17 . 


s 


D 


18 . 


s 


D 


19 . 


s 


D 


20 . 


s 


D 



Name 
Age 



Tape 1 : 



RESPONSE SHEET FOR THE FORCED 
CHOICE PERC E PT UAL PR OC ED UR E 



G r o u p 



Date 



94 



2. 


A 


B 


3 . 


A 


B 


4 . 


A 


B 


5 . 


A 


B 


6 . 


A 


B 


7. 


A 


B 


8 . 


A 


B 


9 . 


A 


B 


10 . 


A 


B 



11 . 


A 


B 


12. 


A 


13 


13 . 


A 


B 


14 . 


A 


B 


15 . 


A 


B 


16 . 


A 


B 


17. 


A 


B 



IS 



19 . 



20 . 



A B 



Tape 2 



2. 


A 


B 


3 . 


A 


B 


4 . 


A 


B 


5 . 


A 


B 


6 . 


A 


B 



/ . A 



A B 

A B 



1] . 


A 


B 


1 2. 


A 


B 


13 . 


A 


B 


14 . 


A 


B 



J 5 



A B 



16 . 


A 


B 


17 . 


A 


B 


18 . 


A 


B 


19 . 


A' 


B 



10 . A 



20 . A B 



95 



Tape 


3 : 




1 . 


A 


B 


2. 


A 


B 


3 . 


A 


B 


4. 


A 


B 


5. 


A 


B 


6 . 


A 


B 


7 . 


A 


B 


8 . 


A 


B 


g 


A 


B 


. 


A 


B 



11 



A B 



12. 


A 


B 


1 3 . 


A 


B 


14 . 


A 


B 


1 5 . 


A 


B 


16 . 


A 


B 


17 . 


A 


B 


IS . 


A 


B 


19 . 


A 


B 


20 . 


A 


B 



Name : 
Date: 



96 



RESPONSE SHEET 
FREE-CHOICE PERCEPTUAL PROCEDURE 

Group: 



Place an A i ; 



an example of category A; place 

Prac tice 



the blank next to the number if the sample is 
i 3 r e a B if an e x a m pie of catego r y B . 



1 . 


1 

t 
c 

1( 

1 tc 

25 
26 

2 7 
2S 
29 
30 
31 
32 
33 

3 4 

3 5 
36 
37 
38 
39 
40 
41 

4 2 
4 3 
4 4 
4 5 
4 6 
4 7 
48 




7 3 . 

74 . 


97. 




2 . . 

3 . 


7 . 




1 . 




4 . 


) . 




5. 


) . 




Ex per i men t al 
1. 


5 m s 

49 . 




2 . 


50 . 


98 . 




3 . 


51 . 


75 . 


99 . 




4 . 


52. 


76 . 


100 . 




5 . 


53 . 


77 . 


101 . 




6. 


54 . 


78 . 


102. 




7. 


5 5 . 


79 . 


103 . 




8 . 


56 . 


SO . 


104 . 




9- 


57 . 


81 . 


1 5 . 




10 . 


58 . 


8 2 . 


106 . 




1 1 . 

12. 


59 . 


83 . 


107 . 




60 . 


84 . 


108 . 




1 3 . 

14 . 


61 . 


8 5 . 


109 . 




62. 


8 6 . 

87 . 


110 . 




15. 

16 . 


63 . 


Ill . 




6 4 . 


88 . 


112. 




1 7 . 

1 S . 

19. 

2 . 
21 . 

22. L_ 

2 3 . 

24. 


65 . 


89 . 


113. 




6 6 . 

67 . 


9 . 


1] 4 . 




91 . 


115. 




6 8 . 

6 9 . 


92. 


116. 




93 . 


117. 




7 . 


9 4 . 


118 . 
119. 




71 . 


9 5 . 




7 2 . 


9 6 . 


1 20 . 





APPENDIX C 

RESULTS OF THE ANALYSIS OF VARIANCE 
FOR SIMPLE EFFECTS FOR ALL PROCEDURES 



98 



Table 17 . Summary of the analysis of variance for simple 
effects conducted on the register/frequency/intensity inter 
action for the vocal i n t e n s i t y d a t a . 



Source 



df 



MS 



RFI Simple 

I n t e r a C t i o n s 



RF @ I 



mm mum 



RF @ 1 



most comfortable 



RF @ I. 



maximum 



2 
2 
2 



80 . 1 

3 . 3 

47 . 8 



4 . 4* 

. 2 
2. 6 



R F .1 Sj in g I e Main 

Effect s 

R © F 1 

25% minimum 

R @ F .1 

2 5% most 



comf ortabl e 



R @ F I 

25% maximum 

R @ F I 

5 0% minimum 

5 % m o s t 



comfortabl e 



R @ F „I 

5 7o maximum 



R @ F I 

7 5 % u i i n i m u m 

R @ F I 

7 5 /o m o s t 

c o m f o r t a b 1 e 

R @ F I 

7 5 % m a x i it u m 



1 
1 

1 

1 
1 

I 



. 7 

769 . 8 
2399 . 5 

20 . 

605.5 
1953.1 

318.5 

770 . 3 
1 2 49 .8 



.0 

37.6* 
117.4* 
.9 

29 . 6* 
95.5* 

15.5* 

3 7.7* 
61.1* 



* S i g n i f i c a n t at t h e . 5 1 e v e 1 



99 



Table 18. Summary of analysis of variance for simple effects 
conducted on register intensity interaction for the vocal 
i n t e n s i t y d a. t a . 



Source 



df 



MS 



F @ R I 

m minimum 



18 . 3 



F @ R I 

m most 

■ comfortable 



F @ R I 

m maximum 



F @ R.I . . 

i minimum 



2 
2 



18 3.9 
78 . 5 
39 . 7 



7 .8* 
3 . 3* 
1 .7 



F @ EI 

* mo s t 

comfort abl e 



F @ R„I 

i maximum 



2 
2 



203 . 4 
287 . 4 



8 . 6* 
12.1* 



♦Significant at the .05 level. 



100 



Table 19. Summary of the analysis -of variance for simple 
effects conducted on the observer group/subject interaction 
(GS) and the subject group/frequency (SF) interaction for 
the paired comparison procedure. 



Source 



df 



MS 



G S Interact!' 



S @ 



n a i v e 



7 . 4 



11 . 4* 



S @ 



ex peri en c ed 



24 . 2 



37 . 2> 



s 1! Intei' acti o n 



S @ F 



20; 



.9 



S @ F 



262 



40 .0 



43 . 0* 



S @ F 



330 



60 .0 



64 . 5 : 



^Significant at .05 level 



101 



Table 20. Summary of the analysis of variance for simple 
effects con cl u cted on t he observ e r groups/subjec t groups/ 
frequency (GSF) interaction for the forced choice percep- 
tual procedure. 



Source 



df 



MS 



GF @ S 



n a. i v e 



4 . 7* 



GF @ S 



singers 



. 1 



1 . 2 



♦Significant at the .05 level 



102 



Table 21. Summary of the analysis of variance for simple 
effects conducted on the subject group/register/frequency 
interaction (SRF) and the register/frequency interaction 
(RF) of the first response data in the free choice procedure, 



Sour c e 



df 



MS 



SRF I nter action 



SR @ F 

208 



SR @ F 



262 



SR @ F 



330 



RF Interaction 



1 
1 
1 



1 . 3 
6 .9 
5 . 5 



1 . 3 

7. 2* 
5.7* 



R @ F 



R @ F 



R @ F 



208 



262 



330 



F @ R 



modal 



F @ R 



f al set to 



1 
1 
] 
2 
2 



24 .9 
1 . 5 

16 .1 
2.9 

23 . 2 



10 .1* 
.6 
6 .6* 
1 .0 
8 . 3* 



♦Significant at the .05 level 



APPENDIX D 
EXAMPLES OF HARMONIC SPECTRUM 



104 



N 

X 
CO 

o 

C\J 
II 



o 
I- 
h- 

LU 
CO 
-J 
< 
Li. 



< 
Q 
O 
2 



o 

(NJ 

+ 



o 



{°i 3tt) SP Nl 



i 

o 

CxJ 
I 



o 
ro 
i 



o 

! 



O 
I 



f 

o 
co 



o 

1 



3GflindlAlV 3AJ±\T13d 



o 




■a 




2 




(!) 




00 




H 




r- 




n 




* J ™ 









CO 




g 




IO 


QC 


(V 






UJ 


,c 




S; 


03 


p 




(O 


s 


a 

•H 




00 


3 








2 


■a 






— 




o 




o 


. - 1 


3 






<T 


t3 

o 




05 


_ 




en 


H 


p. 




t- 


ir 


m 






<i 


C 




to 


a. 



•H 

P 




Ifl 




cd 




V 





£3 




l<> 




ft 




CJ 




Sm 

<H 

>> 

id 




o 




r-l 




CM 




ft 

(0 




CO 








CO 






N 


N 




£ 


K 






3 




CO 




P 


oo 

o 






O 


<N 


in 


LC 







2 


IiJ 

m 


ft 
to 


P 


ro 


**™ 


o 


p 




«£ 


•H 


o 


CJ 


3 





0) 


— 


22 


£ 


A 






Sh 


3 


O 


_J 


td 


to 


OJ 


< 


* 


CD 


CO 


rr 


G 
Cd 


O 


N 


< 





j3 


CO 


a. 










0) 


(0 


«n 




H 


Sh 






ft 


0) 


^ 




td 


P 
(0 


fO 




X 


•H 






w 


bD 


w 






SB 
5-t 






ffi 



P 






CD 


P 






!-i 


CD 






3 


to 






be 


t-H 






•H 


cd 






f*H 


<H 



lUfc> 



N 

X 
C\! 
CO 
C\J 
II 

o 



o 

H 
h- 

LU 

W 

< 
Li. 



O 
N 

• 2? 

CO 
<£> 

_ in 
_ ^ 

- ro 

CM 

o 

0) 
■ CO 



-J 
< 

Q 
O 



to D- 

ro 

CM 



O 

CM 

£ 

CO 

<o 



h- 



o 
+ 



I 

o 

+ 



~r 

o 



o 

CM 
I 



I 

o 

ro 

I 



O 

I 



o 

IT) 
I 



to 

CM 



o 

CTi 
CO 



CO 
10 

to 

J— CM 



O 

CD 

I 



O 
I 



LU 
CQ 

~z 
< 

< 

CL 



d 
a 



■d 

o 
e 



tr 


Xi 


LtJ 


-P 


CQ 


C 


:> 


■H 


Z> 


d 


^2 


CD 

o 


-J 


3 
■0 


< 





H 


Ch 


CC 


en 


< 


G 





cd 
fl 
o 
a 
a 

Sm 
O 
<H 

>» 

cd 

ft 

cn 

•H 

■a 

6 N 

3 K 
!h 

+j CM 

O CD 

CD CM 

a 
en -p 



o 

0) 

£5 
3 
en 


fi 

o 

>> 



(°i3cj)sp ni aanindwv 3Aijlvi3U 



S3 

o 
g 

Cd 
,d 

G 
CO 

<H 

o 

CI) 

H « 

ft U 

E 0) 

cti +-> 

X en 

W -H 

bfi 



• M 

O 

H 
" -M 
CD -P 
M <D 
3 en 
M H 

-H cd 

Pt4 <H 



106 



O 

+ 



o 

!0 



i r 

o o 

ro *r 

+ I I i I 

(°j 3d) sp ni aahindwv 3M±vi3a 



o 

(M 
I 



O 

CD 

i 



< 

to CL 




o 


X) 


0J 


Q 


CD 




* — ' 


H 


CO 


(fl 


~" 


•a 


f- 





— 


e 


to 




■ — " 


0) 


- CC 




* LJ 


B 

■H 


ro CO 






C\J •— » 


T3 



I 
o 

1 



Q) 
O 

x) 

o 
u 
ft 

to 

a 

o 

•H 

cd 

fi 
o 

X! 
ft 

Fh 

O 
<H 

>> 
ctj 

^H 

ft 

CO 
•H 

■O 

£ N 

3 X 
u 

+> o 

o co 

CD CO 

a 
en 






o 



cd 



-p 

(0 

o 

to 


a 
o 

CO 

d) 
■P 

to 

•H 

to 

0) 



o 

■(J 

0) 

to 
l-l 

a 



BIBLIOGRAPHY 



Baisler, P., "A study of in tr al aryngeal activity during 

production of voice in normal and falsetto registers," 
Ph.D. Dissertation, Northwestern University (1950). 

Brown, W. S., personal communication (1969). 

Chiba, R., and Kajiyama, M., The Vowel : Its Nature and 

Structure (The Phonetic Society of Japan), pp., 17-30 
(1958) . 

Damste, P. H., "X-ray study of phonation," Folia Phoniatrica . 
20: 65-88 (1968). 

Donovan, R., "Variables of laryngeal tone," Folia Phoniatrica , 
19 : 281-296 ( 1967) . 

Flanagan, J. L., and Landgraf, L. L., "Self-oscillating source 
for vocal tract synthesizers,". IEEE Transactions on 
Audio and Electro-acoustics . AU-16: 57-64 (1968). 

Hollien, H., and Curtis, J. F., "A 1 aminagr aphi c study of 

vocal pitch," Journal of Speech and Hearing Res earch, 
3 : 361-371 (i960) . 

Hollien, H. ( and Moore, G. P., "Measurements of the vocal folds 
during changes in vocal pitch," Journal of Speech and 
Hearing Research . 3:158-165 (1960). 

Hollien, H., "Vocal pitch variation related to changes in 
vocal fold length," Journal of Speech and Hearing 
Research . 3:150-156 (1960). 

Hollien, H., Moore, G. P., Wendahl, R., and Michel, J., "On 

the nature of vocal fry," Journal of Speech and Hearin g 
Research, 9:245-247 (1966). 

Hollien, H., Coleman, R., and Moore, G. P., "S t r obosc o pi c 
laminagraphic of the larynx during phonation," Acta 
Oto-Laryngologica . 65:209-215 (1968). 



107 



108 

Hollien, H., and Michel, J., "Vocal fry as a phonati onal 
register," Journal of Speech and Hearing Research , 
11 : 600-604 ( 1968) . 

Hollien, H., Personal communication (1968). 

Hollien, H., Dew, D., and Beatty, P., "A study of adult 

phonational range," unpublished manuscript (1969). 

Hollien, H., and Colton, R. H., "Four 1 aminagr aphi c studies 
of vocal fold thickness," Folia Phoniatrica (1969) 
in pre s s . 

H o o 1 e , G . E . , Physiology of the Vocal Registers , Peterson, 
Glasgow ( 1902) , 

Isshiki, N., "Regulatory mechanism of voice intensity varia- 
tion," Journal of Speech and Hearing Research , 7:17 
(19 6 4). 

Kunze, L., "An investigation of the changes in subglottal air 
pressure and rate of air flow accompanying changes in 
fundamental frequency, intensity, vowels, and voice 
register in adult male speakers," Ph.D. Dissertation, 
University of Iowa (1962). 

Ladefoged, P., "Subglottal activity during speech," Proceedings 
Fourth International Congress of Phonetic Sciences , 
The Hague: Mouton, pp. 73-91 (1962). 

Large, J., "An acoustical study of isoparametric tones in the 
female chest and middle registers in singing," The NATS 
Bulletin . December, 12-15 (1968). 

Luchsinger, R., "Physiologie der stimme," Folia Phoniatrica . 
5: 57-129 (1953) . 

Luchsinger, R., and Arnold, G., Voice-Speech-Language: Clin - 
ical Communicology its Physiology and Pathology . 'Bel- 
mont, California: Wadsworth Publishing Co. (1965). 

McGlone, R., "Air flow during vocal fry phonation," Journal 
of Speech and Hearing Research . 10:29 9-304 (1967). 

McGlone, R., and Brown, W. S., "Identification, of break 

between vocal registers," paper EE7 , 75th meeting of 
the Acoustical Society of America, Ottawa, Ontario 
( 1968) . 



10 9 



Michel, J., "Vocal fry and harshness," Ph. D'. Dissertation, 
University of Florida (1964). 

Michel, J., and Hollien, H., "Perceptual differentiation of 
vocal fry and harshness," Journal of Speech and Hear - 
ing Research , 11:440-443 (1968). 

Miller, R . L. , "Nature of the vocal cord wave," Journal of 

the Acoustical Society of America , 31:667-677 (1959). 

Mills, W., Voice Production in Singing and Speaking . Phila- 
delphia: Lippincott and Co. (1913). 

Morner, M,, Fransson, N., and Fant, G., "Voice register termi- 
nology and standard pitch," Speech Transmission Labora - 
tory Quarterly Status and Progress Report , 1:17-23 (1964) 

Murry , T., "Subglottal pressure measures during vocal fry 

phonation," Ph.D. Dissertation, University of Florida 
(19 6 9). 

Nadoleczny, M., "Physiologie der stimme und sprache," in 

Denker, A., and Kahler, 0. ( ed . ) Handbuch der Hals-Nase n- 
O hren Heilkunde (Julius Springer and J. Bergman, Berlin) 
1925 . 

Preissler, W., " St immumganf e gattungen der menschl i chen stimme," 
Arch. Sprach Stimmheilh , 3: 6 5-85 (1939). - 

Ridpatti, R., "The falsetto voice," Laryngoscope . 38:464-471 
(1928) . 

Rubin, H., and Kirt, C. C, "The falsetto: a high speed 
cinematographic study," Laryngoscope . 70:1305-1324 
(1960) . 

Seiler, E., The Voice in Singing . Philadelphia: L. B. Lippincott 
and C o . (1868) . 



Smith, S., "Observations on the vibratory movements of the 

human vocal cords," Folia Phoniatrica , 9:32-35 (1957). 

Sonnesson, B., "On the anatomy and vibratory pattern of the 

human vocal folds," Acta Ot o-L ar yngologi c a , suppl. 156 
(1960) . 



&+*. j+r if*"" 



b& 



3- 2 



110 



Sonninen, A., "Is the length of the vocal cords the same at 

all different levels of singing?" Acta Otolaryngologica 
suppl. 118:219-231 (1954). 

Sonninen, A., "Par at asi sgram of the vocal folds and the 
dimensions of the voice," Proceedings of the 4th 
International Congress of Phonetic Sciences , 
Helsinki, 250-258 (1962). 

Stout, B., "Dynamics of the human voice," Proceedings, Music 
Teachers National Association , 38:299-303 (1944). 

Timcke, R., von Leden, H., and Moore, G. P., "Laryngeal 

vibrations: measurements of the glottic wave, AMA 
Archives of Otolaryngology , 68:1-19 (1958). 

van den Berg, J. W., "Vocal ligaments versus registers," 

Current Problems in Phoniatrics and Logopedics , 1:19-34 
(1960) . 

van den Berg., J. W., "Physiology and physics of voice pro- 
duction," Acta Physio Pharmocol. Neerl .. 5:40-55 (1956). 

Vennard, W., Singing: The Mechanism and the Technic . New 
York: Fischer, Inc. (1967), 

Vennard, W., personal communication (1968). 

Wendahl , R., Moore, G. P., and Hollien, H., "Comments on vocal 
frv." Folia Phoniatrica . 15:251-255 (1963). 

Winer, B. J., Statistical Principles in Experimental Design . 
New York: McGraw-Hill Book Co. (1962). 









BIOGRAPHICAL DATA 



Raymond H. Col ton was born March 17, 1942, at 
Springfield, Massachusetts. He was graduated from Enfield 
High School in June 1959. In June 1963, he received the 
Bachelor of Science degree, with a major in social science 
from Central Connecticut State College in New Britain, Con- 
necticut. From September 1963 to September 1964 he was em- 
ployed by the State of Connecticut as a social worker in 
the Division- of Child Welfare of the Welfare Department. 
In September 1964, he enrolled in the Graduate School of 
the University of Connecticut with a major in speech pathol- 
ogy and audiology. As a VRA trainee, he worked in the Speech 
and Hearing Clinic at the University and at Hartford Hospital 
from September 1964 to September 1965. From September 1965 
to January 1966, he worked as a half-time assistant in the 
Speech and Hearing Department at the Newington Hospital 
for for Crippled Children. From January 1966 to June 1966 
he was a part-time speech therapist at the Unc as -on -Thame s 
Chronic Disease Hospital. He com pie fed all requirements 
for the Master of Arts degree in August 1966 and received 
the degree in June 1967. In September 1966 he enrolled in 
the Graduate School of the University of Florida and has 
pursued work for the degree of Doctor of Philosophy in the 

111 



m>sL 



Communication Sciences Laboratory 



Raymond II. Col ten is married to the former Gloria 



Jean Buonanno and has one daughter Laurie Jean. He is a 
member of the American Speech and Hearing Association and 
the Acoustical Society of America. 



This dissertation was prepared under the direction of 
the chairman of the candidate's supervisory committee and 
has been approved by all members of that committee. It* was 
submitted to the Dean of the College of Arts and Sciences and 
to the Graduate Council and was approved as partial fulfill- 
ment of the requirements for the degree of Doctor of Philosophy 
June 17, 19 69 



^JSUp^^p^o^ 



TV? 



Dean, C o 1 Ljfeg'e ,<5 f A/rts and. Sciences 



Dean, Graduate School 



Super vi sor y C orami 1 1< 



uUU <H> 




\1(W>\ fi\g&0 



6 90 5C